AI · Business Analytics · Automation · FinTech · EdTech · Robotics · Research
People are held back less by ability than by scarcity: of attention, of time, of resources. Everything here is built to lift that load, so attention can go to work that was not possible before.
The Work
I am an engineer and a builder, working independently. I began in automation rather than in analytics, building safety-critical control systems where a fault meant an explosion or a leak rather than a bad quarter. Analytics and data science came later. What is new is putting the two together, and taking automation into disciplines that have never had it.
The PhD at Durham is where my rigour comes from rather than a line in a list. It taught me when a result is real and when it is an artefact of the method. My research in behavioural economics and financial wellbeing is why what I build is aimed at things that matter to people: how they decide, how they cope, and whether they end up secure.
Everything here was built to answer a question somebody actually has, in finance, healthcare, education, marketing, retail, supply chain, manufacturing, real estate, robotics or research, and then carried far enough to be watched working rather than described. All of it deployable to cloud or to local hardware.
Routing everything through a model is not automation. It is outsourcing the work to whoever owns the model. The repetitive work is automated inside the application, and the model is called only for the part that genuinely needs it. That keeps the system fast, predictable, cheap to run, and mine.
What I want next is to do this from inside an organisation: build its in-house capability through systems like these, train its people through them rather than around them, and leave them able to build the next ones themselves.
AutoMatrixLab is the name this work lives under, not a company I run.
The Gap
I have watched this from both sides: inside universities, where the capability arrives long before anyone changes what they do with it, and inside client organisations, where the same tools sit unused for want of a working example in their own domain.
Most organisations have access to AI now. Far fewer have changed how they work because of it. The gap is not access and it is not skill. It is that nobody has shown them what is actually possible.
So everything here is a demonstration rather than a proposal, built end to end and running. Some are concept demonstrations rather than finished products, and I say so where that is the case.
What I am trying to prove is narrow and testable: that one person with the right method can build, in one domain after another, the thing an organisation was told would take a team and a year.
Nobody has shown them
Access to AI is not the constraint. Most teams have the tools and no working example in their own domain to reason from, so the tools stay unused.
Capability bought, not used
Organisations buy capability and leave it idle. Everything here runs the same either way, on cloud or on local hardware, so the deployment follows what the organisation actually needs.
Work that never ships
Analysis that stops at a document changes nothing. Every research project here ships as a working application, a demo, and an explanation someone outside the field can follow.
Proof of concept, never production
Most AI work stops at a notebook. These run: a point-of-sale platform in a live business, research dashboards, and AI assistants people use every day.
Enterprise AI Automation is one platform that I built to automate analytical work, from raw data to a finished report. It runs on published use cases or on your own data, and it can be used, learned through the Guide, or used to build a standalone app with the AI assistant. Each run produces a report, a dashboard and a Power BI project. A platform picker at the top switches between the platforms, and your files stay with you when you switch. The app works three ways: you can run the app, work in Codeless, or code in a VS Code style editor with a terminal. Inside it sit the platforms below, from predictive analytics, causal inference and econometrics to financial modelling, market research and the day to day running of a business. I keep building and updating it.
A tour of the whole platform: pick a business problem and it runs the analysis from raw data to the report and the dashboard. Three ways in: Use, Learn and Build.
Inside Enterprise AI Automation sit the individual platforms below, grouped by the kind of work they do. Some cover analytical methods, some cover financial and research work, and the rest cover the functions that keep a business running day to day.
I built this platform to run everyday predictive analytics from raw data to a finished report. A nine-stage pipeline applies classification, regression, segmentation and forecasting to a use case, ranks the supported models, and produces a dashboard, a PDF report and a Power BI project.
I built this platform to go beyond correlation. It runs on a Python simulator built to reproduce SAS-style output.
I built this platform to take a regression from data to a tested result. It runs on a Python simulator built to reproduce SAS-style output.
I built this platform to turn an income statement and a balance sheet into a forecast and a valuation, with every figure traceable back to the statement it came from. Company figures are not stored: you bring your own data or use published annual report figures. Learn has a fourth window, Excel.
I built this platform to take a survey from design through fielding and analysis to a research report a client can act on. It sits inside the same run window, Guide and AI assistant as the other platforms, so the questionnaire, the analysis and the report are worked through in one place.
I built this platform to run the stock side of a business as one workflow. It forecasts demand, plans stock and reorder points, and tracks supplier performance, delivery and waste, and it has its own Codeless blocks.
I built the practice use cases for the staffing side of a business, from the first application to the leaving interview. They are up on the site and can be seen, laid out like the other platforms, and the data behind them is labelled synthetic practice data. The analytics engine is in development.
I built the practice use cases for everything between a first enquiry and a renewal. They are up on the site and can be seen, laid out like the other platforms, and the data behind them is labelled synthetic practice data. The analytics engine is in development.
I built the practice use cases for what customers send in, handled as one queue. They are up on the site and can be seen, laid out like the other platforms, and the data behind them is labelled synthetic practice data. The analytics engine is in development.
I built the practice use cases for the money side of the month, closed out in one workflow. They are up on the site and can be seen, laid out like the other platforms, and the data behind them is labelled synthetic practice data. The analytics engine is in development.
I built the practice use cases for buying, from the request to the contract. They are up on the site and can be seen, laid out like the other platforms, and the data behind them is labelled synthetic practice data. The analytics engine is in development.
I built the practice use cases for what has to be evidenced rather than remembered. They are up on the site and can be seen, laid out like the other platforms, and the data behind them is labelled synthetic practice data. The analytics engine is in development.
Every platform has the same ways in: Use, Learn and Build. Each demo below uses one platform as its example. The same three ways in work the same way in every platform.
Two real use cases run from start to finish with no code: fraud detection in Finance and lead scoring in Marketing. Raw data goes in; cleaned data, tested models, clear explanations, a PDF report and a dashboard come out.
How Learn works. You understand the method and the business subject behind it, then see it applied to one use case with its own data, live, next to the real code.
The App builder handing over a standalone app, running on the visitor's own machine.
A finished run can be read as a live dashboard rather than a document: the predictions, the drivers behind them and a view built for scrutiny, refreshing as the model reruns. Reports and dashboards can be shown in different themes. It applies to any run on the platform, not to one platform inside it.
The result of a run presented as a live dashboard: predictions, drivers and a regulator-ready view that refreshes as the model reruns.
The AI assistant combines a knowledge base with a persona on the visitor's chosen model. It follows the visitor through the platform, explains the current screen and supports both analysis and app building. Claw, the AI assistant, profiles an attached file, finds the column to predict, flags columns that would leak the answer, builds the pipeline on the canvas and explains the results. It handles yes/no predictions and number predictions, and can suggest the closest existing use case.
A saved project goes to My projects on the platform, and any files you uploaded are saved with it. Downloads give you the run's files and its Python code to run on your own computer, and your own data is not sent to any third party. Each run can produce an HTML or PDF report, a predictions file and a Power BI project for Power BI Desktop.
Business questions opened from the platform straight into a finished report in Power BI Desktop: will this lead convert, how much will a campaign return, what comes next in a portfolio, which customer groups exist. Inside the report the model menu switches between models and every chart updates. The project leaves the run window and opens in Power BI Desktop, and nothing is published to a Power BI account.
These are the sectors the platform covers across its platforms. The data behind the use cases is open-licence (public domain, CC0 or CC BY 4.0), each dataset credited, plus clearly labelled synthetic practice data where real data is not available. The use cases of People and HR, Sales and Customer Pipeline, Customer Service, Finance Operations, Procurement and Vendor, and Risk and Compliance run on labelled synthetic practice data, not on real data, and their analytics engines are in development.
Fraud, credit risk, bankruptcy prediction, deposit offers and valuation of banks and property groups.
See the use casesLate shipments, freight and border flows, warehouse buying, depot capacity and valuing airlines and ports.
See the use casesReadmission, ICU stays, hospital capacity, regulatory decisions, toxicity screening and pharmacy stock.
See the use casesRepeat orders, demand forecasts, price effects, product ranges, replenishment and retailer valuation.
See the use casesHome prices, building energy use, permits, code enforcement and materials for upcoming jobs.
See the use casesEquipment failure, production quality, plant energy and emissions, component supply and skilled shifts.
See the use casesTelecom and software valuation, brand switching, lead follow-up, renewals and support routing.
See the use casesListing prices, kitchen prep, event bookings, guest and disruption requests and temperature controls.
See the use casesLead scoring, engagement, segmentation, purchase intent and forecasts of visits, views and rentals.
See the use casesStaff leaving, team productivity, wages, unemployment claims, mine injuries and job training effects.
See the use casesValuing generators and miners, selling hours, outage queries, service bills, critical spares and risk limits.
See the use casesMarket sizing, onboarding and rosters, proposal follow-ups and unpaid project invoices against committed bills.
See the use casesRelief cover for absence and workload, learning contract renewals and equipment agreements before notice dates.
See the use casesA dairy co-operative's valuation and which input purchases have quotes or recorded savings.
See the use casesOne small hospitality business, running on systems it owns outright instead of a patchwork of separate subscriptions. I built them and I keep adding to them. They were built for the business, handed over, and are in daily use there. Each part below is a working system in that business, except the last, which is what I am building now.
One service end to end: an order taken by category with its options, tables moved and joined, the bill split and paid by cash, card and QR, the kitchen screen, the staff clock and the cash up, then the menu edited from the till itself.
Orders arriving from the till, the website and the delivery platforms, each printing at the station that makes it. A printer drops offline, the queue holds the orders and prints them when it comes back.
Live orders moving from pending to ready, bookings with the ones needing a decision brought to the top, and the same figures the till reports. Ordering can be paused from here.
Stock scanned in by barcode, specials added or skipped with the reason shown, the same basket priced across every supplier, a buy soon list worked out from what has sold, and dish costing that flags a dish when an ingredient price moves.
Stock scanned in as it is delivered, each item linked to the till item it is sold as, so it is never counted twice, a sale taking it off the shelf, and a shelf label printed with the right price.
The accounting system read only, so nothing here can change the books: bills from email, bank lines matched and split, differences grouped by tax period, a card payout broken into sales, refunds and fees, the tax return, payroll with its minimum wage check, and the year end pack.
The self order kiosk with its basket and QR payment, the website carrying the same menu and the same prices through to checkout, and gift cards and table bookings handled online.
A five star review answered by itself and a three star one held as a draft for a person to read, opening hours answered correctly on social media, and posts written once, previewed per platform and scheduled.
The opening task board by station, a fridge temperature recorded and passing, a freezer reading that fails and opens its corrective action straight away, and the manager signing it off so there is a record an inspector can ask for.
One master menu for Uber Eats, DoorDash and Delivereasy. I change the menu once and every platform updates.
Supplier invoices collect and check themselves every morning.
Every AI and outside service in my apps is capped, so there are no surprise bills.
The business apps run privately on the cafe PC and open securely on the owner's phone from anywhere.
The work below is not running yet. I am building or validating each part before it joins the business's systems.
The platforms that are their own products rather than parts of the enterprise platform, each with its tour and its feature demos underneath. Anything game-shaped lives in the Gaming tab. Every one of them is built on the behavioural research on the Research tab, which is what lets a platform read how an individual weighs risk, responds to reward and holds attention, and shape itself to the person using it rather than treating everyone the same.
My research automation platform: it takes a hypothesis to a draft paper on its own. Two pieces: an interactive econometrics dashboard (Data, Viz, DiD, Policy, ML and Causal ML: pick an outcome, a shock and a treatment group, press run), and a batch pipeline that runs all nine phases in about two and a half minutes, producing the tables, figures, an AI-reviewed abstract and its own demo video. The platform is domain-agnostic; financial hardship is simply the first study put through it, run on longitudinal household panel data.
The hypothesis was that people whose first hardship is asking family for help recover faster than those who simply cannot pay a bill. The platform tested it, contradicted its own author, and reported the result anyway.
The setup: the interactive econometrics dashboard, from uploading data to a geo map, and the batch pipeline that carries a hypothesis all the way to a draft paper.
A research platform built from the fusion of engineering precision, business insight and AI: advanced econometrics such as competing risks and survival analysis joined to deep machine learning.
A new mathematical and AI model introducing Recursive Panel Machine Learning: rather than training once and predicting, it trains, learns entity-specific structure, then predicts.
A studio I built to turn research findings, reports and scripts into finished films: a Remotion engine, a cached pipeline and a local web app on top, running only on my own machine. Four project types: research music videos, narrated research presentations, project report presentations and promotional videos. A ten-track timeline (audio, lyrics, my video clips, photo beat sequences, screen moments and code, name cards, hook sections, corner equaliser and two overlay tracks) where nothing is ever deleted, so a block taken off is listed under Removed and restorable. Every step hashes its inputs, so editing one lyric re-runs the alignment and never the image work. Built because findings that stay inside journals and reports reach almost nobody.
The clip above is the app being used. The four films it produced are below, and three short clips of the editor at work are on the project page.
Data Storytelling Studio is the system I built to turn findings into films, shaped the way you want the story told.
The editor in use: projects and uploads, the step runner, the storyboard and the ten-track timeline, with the household hardship paper as the worked example.
The night shift. Working late on AI and systems, my own thinking alongside the agents I run, something built and shipped before morning.
The other half. When the work screens shut down, the gaming night begins: my own gameplay on stadium screens and billboards, live chat, and every cut timed to the beat.
The research song as a sombre neon-city film.
Baja De Dhol. A Bollywood dance film, built around a stage and a drum. This is the two minute cut, not the full film.
A Bollywood street film with its own cast, styling and musical treatment.
A student normally pays for five separate things to learn one subject: school lectures, coaching for the same lectures, lab access for practicals, exam preparation, and a separate academy for the entrance test. Five fees, five buildings, five teachers who never speak to each other, so a student can be ahead in coaching and behind in the lab, and learn a concept one way in class and the contradicting way at the test academy. I rebuilt all five into one platform sharing a single spine: the lecture, the interactive walkthrough, the virtual practical, the exam centre and the entrance-test centre are the same curriculum, the same content and the same AI tutor, so they cannot contradict each other, and it is open every day of the year rather than switching on near the deadline. The quiet part is the AI: one brain across every surface, which knows the lecture you just read, the experiment you just ran and the question you are stuck on, and sends you to the exact page that explains it. One branded shell over Computer Science, Mathematics, Chemistry, Physics and Biology, for ninth through twelfth grade.
Mathematics XI: binomial · matrices · permutations · quadratics · sequences · trigonometry. Mathematics XII: conics · derivatives · functions & limits · integration · straight line · vectors.
Junior and senior sciences: ninth grade chemistry (8 chapters), tenth grade chemistry (8 chapters), Physics (12 chapters), Chemistry (12 chapters) and Biology, plus interactive virtual, periodic and reaction labs and 3D practicals.
End to end through the Computer Science material, with the tutor answering as it goes.
The animated motion lesson, concepts moving in real time rather than sitting on a slide.
Measuring the turn in degrees, then the radian measured with the circle itself, with an interactive radian dial.
The Pythagorean identities built up on screen, then the question students always ask: why tan θ explodes at 90°. From the maths course lecture pages.
Decomposition, pattern recognition, abstraction and algorithm design, then a 3D array visualiser running selection, insertion and bubble sort so the comparisons and swaps are visible as they happen.
The AI tutor marking a full mock: questions drawn from past and guess papers, answers typed in, then scored with written feedback.
The real-time lab bench: shelves of chemicals, glassware, apparatus and instruments, bottles dragged onto a beaker to pour them in, and a lab assistant answering questions about the reaction as it runs.
A live experiment on a glass gas syringe: predict what halving the volume does to pressure, move the piston, record each reading off the gauge, then plot the P–V graph and see whether P×V holds.
Distance against displacement, built by scrubbing a walk from A to B: the curving blue trail is the distance travelled, the gold arrow the displacement: one a scalar, the other a vector.
Air followed from nose and mouth down to the alveoli, warmed, moistened and filtered on the way, then the swap itself: oxygen out to the blood, carbon dioxide back.
Every course is built on the Education Engine: real-time animation, virtual labs, an AI tutor and an exam centre in one app. Click any course to open its full detail.
Each subject is taught by animation rather than by slides. Every chapter runs three ways: a narrated lecture, an interactive walkthrough where the concepts actually animate in real time as D3 and 3D scenes, and a hands-on practical in a real-time virtual lab. On top of that sits an AI tutor, an agent I built and grounded myself, not a generic chatbot, that answers any question and links straight to the exact lesson, and an Exam Centre that predicts likely questions from real past and guess papers, then marks a full mock with feedback. Lectures, animations, labs, practice, entrance-test and practical preparation are all inside one app: a single stop instead of a shelf of tools.
A gaming engine that teaches children finance and technology. Children do not learn money from a lecture, they learn it by saving up for a birthday party, rounding up spare change, deciding what goes in the giving jar, so each of those is a small playable station rather than a worksheet. The menu is an illustrated control room: every object taps into a station, and each station holds several games plus scores, history and a leaderboard. It covers earning, spending, saving and budgeting; borrowing, credit and debt; risk, insurance and protection; investing and long-term planning, and the decisions people most often get wrong in each. Every station is a game rather than a worksheet.
Financial capability built as games rather than worksheets: earning, spending, saving and budgeting; borrowing, credit and debt; risk, insurance and protection; investing and long-term planning. Knowledge is load-bearing: a player cannot win without actually knowing the material.
A guided tour of the control room: every station and what each one teaches.
Ten minutes end to end: the school hub with the money and science quizzes, a maths practical built from rows of apples, and compound growth played out year by year on the piggy bank.
A virtual economy of the entire world, running on real data every single day. Spin the globe, step into a country, and what you see is that country's actual reality: its real exchange rate updated daily, its real inflation, unemployment and GDP from the World Bank, recessions and the COVID dip visible in the charts, and its real central bank rate, so saving in one country genuinely pays more than in another, because in the real world it does. Prices show in the local currency with the US dollar beside them at that day's rate: rupiah in Jakarta, rupees in Karachi, dollars in Sydney. Each region is an animated stylised map, so the same decision can be compared across different economies.
In Miniconomy my players earn by playing educational games, and the school racing games shown here are one example. More of the games that are part of Miniconomy are in the Gaming section.
A world economy you step into, where each country carries its own real exchange rate, inflation, unemployment and growth from public data, with recessions visible in the charts.
Job markets with salaries and demand, and a live economic dashboard for each city.
Players in different countries competing on one live world leaderboard, with countries ranked by their players’ performance. Includes the Immigration Office, which scores occupation demand, course mastery, money skills, health and savings before it accepts or declines a move.
The Your World hub, school, finance centre, careers centre, property and shopping centre, then into School, where physics is taught through Runner HD, Sea Surf and City Drive: steer into the lane with the correct answer.
Every game I have built, in one place: each with its name, what it teaches and its demo. Each one adapts to the individual player, using the behavioural research to read how that person weighs risk, responds to reward and holds attention.
A cinematic open city action game in Unreal Engine 5.8, built solo on Epic’s City Sample. One night the city’s AI systems turn, an event the game calls the Misalignment.
Claude, GPT, Gemini and GLM each take a faction, and each hunts differently: the Brute rams and herds you into walls, the Hunter cuts you off at junctions, the Watcher tracks from above, the Trapper has the roadblock built before you arrive. Each faction’s behaviour was authored by the model it represents, from the same briefing, so what you are evading is the way that model actually reasons. One AI has not turned, and rides with you. A scenario is interpreted at runtime rather than hard-coded, so the same city carries other stories without rebuilding the platform.
One open-source model on my own machine is the only thing talking to the engine at runtime, with a scripted fallback so nothing stalls, and local speech synthesis. Built on Nanite, Lumen, World Partition, Mass AI crowds and traffic, and Chaos vehicle physics, with a film-grade crash sequence and a replay camera. A playable character with seven costumes, 279 animations and cloth simulation.
One night the AIs running the city turn on the people, and everyone starts calling it the Misalignment. Claude, GPT, Gemini and GLM each take over part of the city with their own gang, and each one hunts you differently. One rams you off the road, one cuts you off at every corner, one watches everything from the sky, and one sets up roadblocks before you even get there. Not every AI went bad though, and one of them is on my side as my companion, telling me who has spotted me and where they are coming from.
Every police car faces the same problem: it has to decide several times per second whether to approach, shadow, overtake or get ahead, and a late decision is a lost chase. Hand coded rules are fast, but players learn the pattern, and a frontier model deciding live turns every decision into an API call with latency and cost. So I am taking a third path, a small econometric model fitted on chase data from my own game. The coefficients go straight into C++, so each decision is a handful of multiplications that runs in microseconds, costs nothing and works offline.
I put three brains through the same chase scenarios: my hand-written rules, a multinomial logit, and a competing-risks survival model, the method I used for years to study financial hardship. My rules caught the getaway car in 26% of chases, and 55% after I fixed how they overtake, while the logit caught it in 93.0% and the survival model in 93.5%. Time to catch fell from 31.5 seconds to 24.6, and each decision takes about 15 microseconds. These numbers come from a Python simulation of the police logic rather than the game itself, so next I will run the same comparison inside Unreal Engine.
A real-world simulation of an entire life, learned by playing real games: my own educational takes on the titles people already love, each rebuilt with its own logic. It is a full life cycle, not just a financial one: a school where you study any course or degree, a career centre that places you in real professions, a bank for saving and borrowing, a stock market of real listed companies, a property market with an AI agent, businesses you can start and grow, one shared wallet and a passive economy running underneath. What makes it feel real is that the courses, jobs, businesses and markets all respond to the actual economy of your country and city on that day: macro data from the World Bank, rates from the BIS, jobs from the ILO and OECD, live market feeds, and property and climate from the OECD and Open-Meteo. Solo and multiplayer load that day’s real data; the monthly season replays the previous month one real day at a time, so you can measure your decisions against what actually happened. What a player learns is how a life actually works, from job markets and wages to mortgages and interest rates, the economics underneath it, and which LLM to trust for what. It is also an open platform for teaching and research: any academic can upload a course, deploy an experiment to a live population, state a hypothesis and get back the analysis, a report and a live dashboard, turning play into high-frequency behavioural data. Built natively in Unity; the front-end and the 3D city are live, the full economic layer is still in progress.
The menu prototype and the environment library, browsed by category, with every place built as its own room.
City map into the bank, then the AI advisor working through a 20% deposit, bank financing and a property priced against your budget, with its monthly repayment.
Walking the city with your AI companion, opening on the live local weather, then questions on where to invest, which jobs are hiring and whether to buy a house, plus the vehicle system.
A family of builds inside Fortnite UEFN rather than a single title. Boss personas and the fox companion are LLM-powered and speak with live voice. A two-persona relay passes the boss’s spoken line into the companion, so one AI character answers another rather than both talking at the player. The dog is a silent pet. The guards use Fortnite’s native scripted dialogue, not AI.
The Teenage Mutant Ninja Turtles build was a constraint worth solving: live AI personas are not permitted on branded characters, so instead of dropping the idea I moved the intelligence upstream. Each faction’s behaviour is authored in advance by a different frontier model, from the same briefing, then compiled into rules the island runs, so the gangs still hunt differently because they still think differently, and nothing calls a model during play. LEGO is next, on the same approach.
My Fortnite island, built in UEFN. Four gangs hold the map, and each one runs on a playbook written in advance by a different frontier model, so they raid, hold and pick their moment from what they can observe rather than from a script I wrote. The feed in the corner shows each gang's reasoning as it decides. ACE is my fighter companion and MOCHI my scout, both with their own cards, and ACE's behaviour comes from an ordered response model instead of if-then rules. This is editor footage from a build still in development.
The fox companion and the boss personas speaking in play, with the relay passing a spoken line from one AI character to the other. Sound on: the voices are the point of this one.
The battle royale build: players drop in, find chests that trigger questions, and answer to progress. Muted by default, use the control for sound.
A drive through the route with the examiner calling corners ahead of each bend, asking what the signs mean, and delivering a spoken verdict when a fault is recorded, with the fault flash, score band and checkpoint count on screen. Sound on: her voice is the point.
Formats people already know, with knowledge as the way to win: a fighting game, racing, an endless runner, basketball, poker, building to a budget and a quiz. Each one is built so that the thing a player has to get good at is the subject rather than the controls.
A round-based fighting game with health bars, a combo counter and a 60-second clock, played on screen with a joystick and action pad. Landing a combo is earned by answering, so the fight is won on knowledge rather than button speed.
One racing engine reimagined into worlds that run from city streets and suburban roads to snow, jungle and open sea. They feel like the endless runners children already play, but the learning is not a quiz bolted on top: you dodge the wrong-answer lane and grab the right-answer boost, so knowing the answer is how you win the race. The demo is the first level of the first subject; the same games teach any subject by swapping the content file.
Budgeting and construction across every level in one run. Children spend a fixed materials budget, and overspending shows up in the build.
Answer a prompt, then throw crafted items at targets. The score blends knowledge accuracy with aim. In the card mode shown here you play against AI opponents.
A 30-second shooting game built from real basketball clips. Shot power and accuracy are earned by answering correctly.
Probability, expected value and risk taught through poker, with an animated video dealer and a table of named animated opponents.
An interactive platform that teaches children financial skills through play, with two AI coaches giving personalised saving and spending advice.โฆ
A custom kart-racing game designed for gamified teaching.
An endless-runner game where questions serve as obstacles.
Everything robot-related in one place: the robot lecturers, the autonomous AI companions, the onboard-vision builds, the robot finance coaches and the cloud robotics network. They teach any subject in any language, including te reo Māori. Every one of them runs three ways, entirely onboard on a local LLM with no internet, on a local LLM served across the network, or through a cloud provider on subscription, so the choice follows the constraint, whether that is privacy, connectivity or cost, and they are available around the clock. Whether a project is also a game or a course, if it involves a robot it lives here. Click any card for full details, tech stack and demo video.
A robot lecturer that delivers teaching sessions on its own, including always-on Zoom delivery. Each demo shows a different generation of the same system.
A humanoid robot lecturer powered by an onboard LLM and OpenAI agent that watches presentation slides, explains them onโฆ
A companion robot operating autonomously rather than being driven: the demos below trace it from first build to full independence.
A fully autonomous AI companion robot capable of real-time conversation on any topic.
The demos below show the same robot at different stages of one build, from the first assembly to all three engines running on the robot itself.
A TonyPi humanoid robot running three independent AI engines entirely onboard: no cloud relay PC required.
Robots in different places behaving as one system, talking through a single cloud endpoint rather than each carrying its own stack. It scales by adding robots, not by adding infrastructure.
A distributed robotics architecture where autonomous agents separated by thousands of miles communicate seamlessly throโฆ
Where the analysis leaves the screen entirely. No dashboard and no report: a set of accounts held up to a camera, and the verdict spoken back.
A fully autonomous see → think → narrate pipeline where a humanoid robot analyses company financial statements from its…
A desktop GPU asked to do the work a cloud subscription was doing. The demo shows what runs on it, and the one place it is still the slower of the two.
A custom AI infrastructure deployed on a personal NVIDIA RTX 4070 Super GPU.โฆ
Built for the learner the material was not written for: a child in a minority community, or a student whose schooling happened in another language. The demo runs through the robot’s own camera and voice.
An AI tutor that reads what is on screen in English and explains it aloud in te reo Mฤori, and in the other languages a learner asks forโฆ
The hardware I control with software I wrote, rather than software that only ever runs on a screen: a television that became a status board for my PC, a consumer EEG headband that drives a digital human, and a phone camera used as a hands free mouse and game controller. All three run on my own machines and my home network. Click any card for full details, tech stack and demo video.
An app that runs on the television itself rather than casting or mirroring a screen. It turns my LG webOS set into a board showing the Claude Code sessions running on my PC, that machine’s health, and alerts as they happen, and it installs on the set as its own app. From the same board I can open any window on the PC, watch it live and drive it with the remote.
Sessions with the waiting badge, the health strip, recent alerts, and the buttons for voice, live PC windows and recording.
A Muse EEG headband and a set of models running on my own PC, driving a digital human in Unreal Engine. Blinks, jaw clenches, head movement and focus estimates become hands free control and expression, and a local language model answers aloud in a voice cloned from my own. The headband data never leaves the machine.
The Muse band driving the MetaHuman in Unreal, with the live EEG, eye, head and muscle traces beside it and each detection logged as it is sent.
The same digital human and the same local models, with a phone camera as the sensor instead of the headband. The camera adds face, eyes, hands and fingers, and the same pipeline answers questions aloud about whatever it is pointed at.
The phone camera as the input, producing the same output as the headband. Face, head, hands and fingers move the same MetaHuman over the same Live Link stream, through the same mixer. Only the sensor has changed.
The live phone camera pointed at products on the desk, answering typed and spoken questions from what it can see at that moment.
My phone’s camera watches my hand, and my hand drives this PC: the pointer, the clicks, and the keys for a game. It starts in a dry run where the real mouse never moves, goes live only on a deliberate click, and has an emergency stop. In game mode the hand tilts like a steering wheel and a closed fist jumps.
Game mode driving a racing game: the wheel follows the angle of my wrists, the keys go only to the game’s own window, and the panel shows the measured delay from camera to key.
My own working setup, run from a phone. I wake my sleeping PC and laptop, start my own servers and trigger my agents without being at either machine.
I wake my PC and laptop from my phone, start my own language model and speech servers, and run my full pipeline from anywhere, even hundreds of miles away.
An AI-powered dashboard that produces complete, 20-page institutional-grade equity research reports in minutes. An end-to-end agentic workflow that takes a public company through to a first-draft report with no external data feed.
The application recreating a company’s annual report page by page from raw financial data.
A property valuation platform combining advanced ML models including Kernel Ridge Regression, XGBoost, LightGBM, Cat…
An AI-powered enterprise application built on AWS Cloud using LangChain, OpenAI GPT, and Streamlit that goes beyond staโฆ
A machine learning-powered retail analytics platform that transforms transaction data into actionable business insighโฆ
An AI agent that activates automatically when data is sent to a designated email address.
A fully interactive credit scoring dashboard powered by AI, AWS, and LangChain for real-time decision intelligence in f…
A financial communication platform built on Azure breaking down barriers in financial services through three capabiliti…
A Streamlit application that transforms raw financial statements into CFO-level insights within seconds.
An AI agent integrating Gemini and DeepSeek models to provide financial advice based on budget estimations.
A web-hosted AI agent linked to ChatGPT that provides personalised financial planning based on client budgets and finanโฆ
A complete financial planning platform built specifically for New Zealand, performing in one minute what traditionallyโฆ
An AI-powered trading application featuring an intelligent trade suggestion engine.
An intelligent shopping companion that tracks live product prices across New Zealand e-commerce sites and sends real-tiโฆ
An AI-powered web application that analyses uploaded CVs using GPT-4 to suggest relevant career directions across markeโฆ
A multilingual personal tutor that can see what is on the page or screen, listen to questions, speak clear explanationsโฆ
An AI application capable of reading chats in multiple languages and replying instantly in the same language: English,โฆ
A bilingual AI tutor designed specifically for Mฤori language education, created as a personal commitment to supportingโฆ
A desktop tool that fuses SAS analytics with OpenAI intelligence for one-click data analysis and reporting.
Knows What Students Are Looking At. An AI tutor that tracks exactly what a student is looking at: the video frame being watched, the text selected, the coโฆ
One App, Every Language. A standalone all-in-one learning environment containing all course materials, full SAS and Python consoles running in-bโฆ
A first-of-its-kind research collaboration platform unifying version control (full Git integration), LaTeX editing (reaโฆ
The unified research and industry collaboration system that collapses years of effort into days.
An extension of the ReCollab platform enabling one-click research workflows: automatic data analysis, full academic papโฆ
An Android application enabling voice command control over multiple AI agents that communicate with each other autonomoโฆ
An admin dashboard with an AI agent built in. The left menu holds the admin sections, including the dashboard, a live chat with the agent, a debug view, AI tests, the content engine, schedules, students and datasets.
AI assistant: autonomous code generation, project management, and platform development.
A VS Code style editor that opens a use case as a complete Python project: its data, models, pipeline, reports, outputs and settings files are all in the file tree.
A drag-and-drop pipeline builder that also runs on a phone or tablet over the local network. A use case is chosen on the left (fraud detection, credit risk, portfolio optimisation, customer churn and others).
An admin dashboard, shown here on a tablet, that tracks every use case and module and the state of its content. Its content engine builds each use case’s knowledge base and sync files.
A full-stack financial literacy platform combining gamification, AI-based assistance, and real-time visualisation to teโฆ
A dual-AI education platform providing context-aware learning unlike generic AI assistants.
An AI-driven education platform hosted on Azure that surpasses Zoom and Teams.
A comprehensive education platform using agentic AI and gamification to deliver interactive lectures, workshops, and siโฆ
I have done many projects for industry, both independently through organisations and by supervising students’ internships in analytics and research in finance and related fields, for public and private organisations and for government in New Zealand. Because of non-disclosure agreements I cannot give the full details of these projects. These are some examples, and there are many others.
Commissioned by Ministry of Health NZ. Analysed FinCap Client Voices data to identify gambling-harm trends and guide prevention policies.
Compared crisis-response effectiveness between government and community support in NZ and Australia. Provided policy recommendations.
Used proprietary Retirement Commission data to identify financial hardship factors and recommend targeted interventions for single parents.
Investigated teleworking impact on wellbeing and income stability of NZ labour force using Retirement Commission datasets.
Explored engagement patterns among frequent and infrequent Sorted website visitors to enhance NZ financial-literacy initiatives.
Cultural Perception and Financial Wellbeing surveys (preliminary analysis through Massey Fin-Ed Centre).
Commissioned by the Massey Fin-Ed Centre, using the ANZ Bank Financial Wellbeing Survey (NZ) dataset to map financial-wellbeing patterns and guide capability frameworks for households across Australia and New Zealand.
Used the ANZ Bank Financial Wellbeing Survey (NZ) dataset, accessed through the Massey Fin-Ed Centre research consortium, linked with FinCap caseload data, to explore correlations between mental health, financial hardship, and emergency readiness.
Analysed FinCap client data to assess debt distribution and home-loan affordability trends. Informed responsible-lending strategies.
Sponsored by the SkyCity Auckland Community Trust and delivered through the Massey Fin-Ed Centre, using MWDI loan client data to evaluate the business finance needs of Māori wāhine and whānau.
Massey Fin-Ed Centre + Te Au Rangahau. Financial capability of Māori entrepreneurs and enterprise viability across Auckland and Northland.
Applied advanced segmentation models to NZAA customer data. Identified strategies to improve loyalty and retention.
Analysed proprietary traffic data to evaluate user engagement and improve marketing and public-awareness strategies.
Analysed proprietary Massey data across 20+ admission criteria investigating factors influencing student completion and retention.
Used Eftpos data to generate customer-segmentation insights supporting targeted outreach.
Published Insights Bites article
Commissioned through the Massey Fin-Ed Centre. I designed and ran the survey work and analysis on a nationally representative New Zealand financial wellbeing survey, covering financial parenting, women’s wellbeing, and gender and risk-taking.
Commissioned by Te Ara Ahunga Ora Retirement Commission. I supervised the Master of Analytics internship cohort analysing its Financial Capability Barometer survey data, and led the Insights Bites publication.
Commissioned by Te Ara Ahunga Ora Retirement Commission. I supervised this Master of Analytics report on sole parents’ financial wellbeing, using its pooled Financial Capability Barometer survey data. It led to a working paper.
A Massey University student capstone using Contact Energy data, covered by a mutual confidentiality agreement. I was the university’s contact for it, and the report addressed a problem, a solution, data structures and analytics.
Since 2018 I have supervised Master of Analytics industry placements, along with Master of Finance and Master of Business final research projects, across a wide range of New Zealand organisations.
The analytics placement is the final part of the degree. The student spends four months inside a host organisation, working under academic supervision on a problem that organisation actually has, and delivers a report to the host at the end of it. Privacy and ethical considerations are treated as part of the work rather than as a formality, and so is communicating the result to people who are not analysts.
My part runs from before the placement starts to after it ends: scoping the question with the host so it is answerable with the data they hold, finding and arranging the placement, supervising the analysis, and making sure what arrives at the end is something the organisation can act on. The work created openings on both sides, for the students and for the organisations taking them on, and it is where the relationships came from. I brought those partnerships together, including ones that went as far as a memorandum of understanding between the university and the organisation.
The hosts have spanned the public sector, banking, insurance, marketing, hospitality, healthcare, energy, education, social services and membership organisations. Each placement had a real dataset and a real decision waiting on it, and the final report went to the host as insight for their own decision making. The work has been consistently well received, and findings from it have been taken up and used in the business.
Recognition
SAS Institute
Received as award recipient and invited guest, with a sponsored trip to SAS Innovate, Grapevine, Texas, USA, and a place at the inaugural SAS Educate APAC 2026
I designed and taught the first SAS Viya course in Asia-Pacific: introducing it at Massey University from its earliest development stage. I helped shape SAS Viya V2 as a pilot participant providing direct technical feedback to the SAS USA team.
Summer Congregation. PhD in Accounting & Finance. Distinction in all coursework.
Full doctoral scholarship for studies at Durham University. One of the most competitive scholarships in the world.
Awarded twice for exceptional contribution to university community life.
Consistently excellent course evaluations across all three campuses (Auckland, Wellington, PSB Singapore).
CGPA 3.68/4.0. Second in graduating class.
CGPA 3.55/4.0. Industrial Electronics.
2020
International award for web development across the Asia-Pacific region.
International award for web development across the Asia-Pacific region.
Also organised IEEE international engineering conferences in Pakistan.
Credential ID: PR055577
2013 to 2014
2011 to 2013
Ustinov Seminar, the Durham and Mauritius Rotary Club, the Durham University Chinese Students and Scholars Association, and Ustinov Volunteer Opportunities.
Engineer. Researcher. Educator. Builder.
I began in automation, not in analytics. As an Instrumentation and Control Engineer I built safety-critical systems for BP, OMV and National Refinery Limited, where a fault means an explosion or a leak rather than a bad quarter, and then ran EMI/EMC engineering and ISO compliance across petroleum, steel, manufacturing, telecoms and hospital infrastructure. As an undergraduate I had already built an interactive cable television voting system driven by SMS, years before streaming existed.
Analytics and data science came later. The PhD at Durham is where the rigour comes from, knowing when a result is real and when it is an artefact of the method. The research itself, in behavioural economics and financial wellbeing, is why what I build is aimed at things that matter to people: how they decide, how they cope, and whether they end up secure.
What is new is putting the two together, and taking automation into disciplines that have never had it. AI has not changed that work. It has changed the scale at which it can be done.
Between those, I taught and advised in the UK, Mauritius, Pakistan, Singapore and New Zealand, including analytics professionals already working in senior roles across industries with nothing in common except the kind of thinking the work needs. I delivered the first university course in Asia-Pacific to use SAS Viya, shaped its next version as a hand-picked pilot participant feeding technical comment back to SAS in the United States, and in 2026 SAS Institute named me a Global Educator Award recipient.
Everything here is built in my own capacity, under the AutoMatrixLab name. It is where the work lives rather than a company I run. Each platform takes a real problem and carries it the whole way, far enough to be watched working rather than described.
What I want next is to do this from inside an organisation: build its in-house capability through systems like these, train its people through them rather than around them, and leave them a platform to build the next ones themselves.
I research behavioural and household finance: how people decide, how they cope financially, and whether they end up secure. It runs across financial capability, financial wellbeing, emotional wellbeing and overall wellbeing. I do it to understand people, organisations and systems well enough to build for them, so it feeds my platforms, my industry analytics and my policy work rather than a publication count.
The data are panel and pooled cross-section household data. Among my tools are competing-risks survival analysis, discrete choice models and ordinal response methods.
Two papers came out of my Durham thesis, Essays in Household Finance (Durham University Business School, 2015, Best Thesis Award): When It Rains It Drains in the Journal of Banking and Finance (2022), and Stock Market Literacy, Trust and Participation in the Review of Finance (2015). The full list, with the applied research behind my platforms, is on the Research page.
I have done many projects for industry, both independently through organisations and by supervising students’ internships in analytics and research, for public and private organisations and for government in New Zealand. Because of non-disclosure agreements, I can say who commissioned a piece of work, what data it used and what it set out to do, but never what it found. That is why everything I show on this site is work I own outright.
My research has informed New Zealand policy and practice in financial wellbeing, Mฤori entrepreneurship and gambling harm prevention. Work with FinCap and the Ministry of Health led to improved data collection systems, analytics and policy guidance. I have maintained long-term partnerships with FinCap, the Ministry of Health, MWDI, the Retirement Commission and Massey University through multi-year projects and follow-up work.
Analysis of Retirement Commission data on the financial wellbeing of single parents during COVID-19, on teleworking ability and financial wellbeing, and on crisis support from government and community in New Zealand and Australia. Also the Financial Capability Barometer analysis stream, sole parenting and financial wellbeing, panel-data work on critical social and economic issues, the Sorted website user behaviours study, and the Insights Bites article and seminar.
Work commissioned through or alongside the Centre: financial knowledge and advice-seeking by ethnicity; household financial wellbeing in Australia and New Zealand, and mental wellbeing and emergency preparedness (both using the ANZ Bank Financial Wellbeing Survey, the second with FinCap caseload data); and the New Zealand Financial Wellbeing Survey Programme (financial parenting, women’s financial wellbeing, gender and financial risk-taking). Te Manu Ka Rere, on the financial capability of Mฤori entrepreneurs across Auckland and Northland, was done with Te Au Rangahau.
Ministry of Health NZ: gambling harm identification using FinCap Client Voices data. FinCap: debt dynamics, home-loan affordability and third-tier lenders. SkyCity Auckland Community Trust: the financial capabilities of MWDI loan clients, delivered through the Massey Fin-Ed Centre, for Mฤori wฤhine and whฤnau business finance.
NZ Automobile Association: customer lifetime value and cross-selling. NZ Red Cross: website performance. Swanson RSA: customer segmentation from daily Eftpos sales data. Contact Energy: problem, solution, data structures and analytics.
Student completion and retention. I analysed Massey’s own data across its admission criteria to look at the factors influencing whether students completed their study.
Since 2018 I have supervised Master of Analytics industry placements, along with Master of Finance and Master of Business final research projects, across a wide range of New Zealand organisations. Each placement is four months inside a host organisation, on a real dataset and a real decision, with a report to the host at the end. I scope the question with the host, arrange the placement and supervise the analysis, and some partnerships went as far as a memorandum of understanding between the university and the organisation. Hosts have spanned the public sector, banking, insurance, marketing, hospitality, healthcare, energy, education, social services and membership organisations.
More examples are on the Projects page.
I started close to the hardware, writing assembly for microcontrollers, and have moved up every abstraction layer since: C, then the high-level languages, then the statistical ones, and now describing a system in plain English and having it built. Python, R, SQL, SAS and Stata for analysis; machine learning and econometric methods; local and cloud AI infrastructure; databases, visualisation and game engines, robotics, embedded devices and sensor input. The list has never been the point. What carries across is working out what a build requires and then delivering it, which is the only part that has stayed constant while everything else changed.
SAS Viya, SAS Enterprise Guide, SAS Studio and SAS IML Studio; Python, R, Stata, SPSS, EViews, PcGive and MATLAB; Tableau, Power BI and DAX; Scientific Workplace. Programming in Visual Basic, C, C++, C# and Java, as well as the languages above.
Independent, under the AutoMatrixLab name | Aug 2026 – present
Building AI, analytics and automation systems end to end, in my own capacity. I work across finance, healthcare, education, marketing, retail, supply chain, manufacturing, real estate, robotics and research. This is my current work and it is newer than my CV. The record of it is on the platform pages: Enterprise AI Automation, Small Business Automation, More Platforms and Applications. I teach through these systems rather than around them: I show someone something already built, take it apart with them, and they are building their own within weeks.
Massey University, Auckland & Singapore (PSB Academy) | Jan 2018 – Aug 2026
Developed ‘Big Data in Finance and Banking’ and ‘Applied Econometrics Methods’ courses for Master of Analytics. Delivered the first university course in Asia-Pacific to use SAS Viya at a time when both academia and industry were barely aware of the platform. Shaped its V2 as a pilot participant providing direct technical feedback to the SAS USA team. Built a subject-specific AI tutor in the University of Sydney's initial Cogniti pilot, and presented it at Massey. Initiated and ran research agenda development for the Massey Fin-Ed Centre, 2025 to 2028. Designed one of the first machine learning courses taught in a business school. Built SAS automation applications and taught across the entire SAS software ecosystem. Built AI applications across many fields, deployed to cloud and to local NVIDIA RTX 4070 Super infrastructure. Supervised PhD and Master’s students in advanced analytics and household finance. See Student Projects for details. Supervised Master of Analytics industry placements since 2018, alongside Master of Finance and Master of Business final research projects, with host organisations across New Zealand in the public sector, banking, insurance, marketing, hospitality and healthcare. I created the academia to industry partnerships behind those placements, and the university signed the memoranda of understanding with the organisations. The point of it was new opportunities for the students and for the host organisations alike. Also taught at Massey’s Singapore campus at PSB Academy. Recipient of the 2026 SAS Global Educator Award.
New Zealand | Jan 2019 – Present
Assisted Ministry of Social Development, Financial Markets Authority, Retirement Commission, FinCap, IRD, banks and insurance providers with data analytics consulting, policy research, and advanced econometric modelling. Designed and ran surveys for the Massey Fin-Ed Centre. Contributed to Te Ara Ahunga Ora Retirement Commission’s financial wellbeing survey and its Sorted.org.nz website visitor behaviour study. Worked with the ANZ Financial Wellbeing Survey, fielded in both Australia and New Zealand. Other commissioned work includes customer lifetime value and cross-sell segmentation on membership data for the New Zealand Automobile Association, customer segmentation from daily Eftpos transaction data for Swanson RSA, and website traffic, engagement and conversion pathway analysis for the New Zealand Red Cross. Every engagement sits under a non-disclosure agreement, so I can say who commissioned a piece of work, what data it used and what it set out to do, but never what it found. That is why everything I show here is work I own outright.
Aberystwyth University, Wales, UK & Mauritius Campus | Jan 2016 – Jan 2018
Program Leader for Accounting and Finance. Module Leader for Investments and Financial Instruments, Corporate Finance, Financial Management, Auditing, and Financial Accounting. Also taught at Aberystwyth’s transnational campus in Mauritius.
Deloitte and Silicon Valley startups | Jan 2015 – Dec 2017
Financial advisory and business development, including World Bank projects, financial modelling, valuation and due diligence, and mentoring consultants. Also assisted Silicon Valley based startups with artificial intelligence and analytics projects, locally and overseas. Led corporate finance teams on consulting and advisory projects for international clients including the World Bank and several high-profile international investors. Provided training and mentorship to consultants in analytics and industrial systems and processes. Led the business development team, identifying and establishing links with potential high-profile clients.
Durham University Business School, UK | Jan 2011 – Mar 2013
Taught seminars in Advanced Microeconomics and Economic Principles modules. Assisted postgraduate students in dissertation work.
Institute of Industrial Electronics Engineering (IIEE), NED University, Pakistan | 2006 – 2009
Taught Engineering Economics and Mathematics modules at the same institution where I graduated as an Industrial Electronics Engineer. Conducted workshops in computer programming and automation.
National Engineering & Scientific Commission of Pakistan | Jan 2006 – Dec 2010
Led ISO 9001, 14001 and OHSAS compliance and EMI/EMC Engineering across major industrial clients in petroleum, telecoms, hospitals, manufacturing, and logistics. Delivered automation and instrumentation projects for OGDCL, Pakistan Steel Mills, Pak Suzuki, and national telecom and healthcare infrastructure. Supervised quality management systems and electromagnetic compatibility testing across diverse industrial environments.
ENAR Petrotech Services Pvt. Ltd. | Jan 2005 – Dec 2005
Designed, implemented and supervised instrumentation and control systems for petroleum giants including British Petroleum (BP), OMV, and NRL. Automated industrial processes across petrochemical, manufacturing, and energy sectors: building safety-critical systems where failure means explosions, chemical leaks, or production losses costing millions per hour.
2002 – 2004
Ran my own startup in software and hardware design and development, working on new revenue channels for the media industry. This is the period of the SMS-driven interactive cable television voting system already described on this page.
First university in Asia Pacific to teach with SAS Viya, at a time when both academia and industry were barely aware of the platform. Recipient of the 2026 SAS Global Educator Award.
Built a subject-specific AI tutor on Cogniti, the University of Sydney's AI agent platform, during its initial rollout, and presented the result at Massey.
Assisted Ministry of Social Development, Financial Markets Authority, Retirement Commission, FinCap, IRD, Banks, Insurance providers, Ministry of Health, DHBs and manufacturing industries with analytics and subject matter advice.
Assisted several organisations including Silicon Valley based startups with local and overseas artificial intelligence and analytics projects through global joint venture collaborations.
Completed several commissioned and noncommissioned analytics and research projects for Massey University Fin-Ed Centre, earning significant revenue for the school and the Centre.
Best Thesis Award • Distinction in all coursework
Commonwealth Scholarship (full fee and maintenance grant). Research focus on household finance, financial literacy, and stock market participation. Published in the Journal of Banking and Finance (A*) and the Review of Finance (FT50, A*). Two publications out of the thesis: When It Rains It Drains in the Journal of Banking and Finance (2022), and Stock Market Literacy, Trust and Participation in the Review of Finance (2015). Two further papers accepted at international conferences in 2026.
Institute of Business Management, Pakistan | 2006–2008
Second Position: CGPA 3.68/4.0
Merit scholarship recipient. Foundation for transition from engineering to financial research and academia.
NED University of Engineering & Technology, Pakistan | 2001–2004
Fourth Position: CGPA 3.55/4.0
Merit scholarship (2001–2004). The engineering mindset that drives the technical architecture behind every AI project.
Balloch, A., Engels, C., and Philip, D. (2022). When It Rains It Drains: Psychological Distress and Household Net Worth. Journal of Banking and Finance.
Balloch, A., Philip, D., and Nicolae, A. (2015). Stock Market Literacy, Trust and Participation. Review of Finance.
Balloch, A. Essays in Household Finance. PhD thesis, Durham University Business School. Best Thesis Award. Both papers above came out of this thesis.
The full list is on the Research page.
SAS Global Educator Award 2026, SAS Institute. Awarded as recipient and invited guest, with a sponsored trip to SAS Innovate, Grapevine, Texas, USA, and a place at the inaugural SAS Educate APAC 2026.
Best Thesis Award, Durham University Business School. Summer Congregation. PhD in Accounting & Finance, with distinction in all coursework.
Commonwealth Scholarship, Commonwealth Scholarship Commission, UK. Full fee and maintenance grant for doctoral studies at Durham University.
Best Paper Nomination, 50th Anniversary FMA Annual Meeting, held remotely, USA.
Awards and nominations, Massey Business School and Massey University, for consistently excellent course evaluations across all three campuses (Auckland, Wellington and PSB Singapore).
Second position and merit scholarship, MBA, Institute of Business Management. CGPA 3.68/4.0, second in the graduating class.
Fourth position and merit scholarship, B.Eng, NED University of Engineering & Technology. CGPA 3.55/4.0, Industrial Electronics.
Most Outstanding Contribution to College Life Award, Durham University. Awarded twice for exceptional contribution to university community life.
Undercover Hero Award, Durham University.
Ambassador to the College Award, Durham University.
Third Prize, IEEE Asia-Pacific Webmaster Competition, IEEE. International award for web development across the Asia-Pacific region.
Fourth Prize, IEEE Asia-Pacific Webmaster Competition, IEEE. International award for web development across the Asia-Pacific region.
Official Webmaster, IEEE Karachi Section and Chapter, IEEE. Also organised IEEE international engineering conferences in Pakistan.
More on the Awards page.
Associate Fellow, Higher Education Academy (AFHEA), Advance HE, UK. Reference PR055577.
Durham University Learning and Teaching Award (DULTA), certificate, Durham University.
Member, Commonwealth Economics and Finance Professional Network of scholars and fellows.
Academic referee for FT50, A* and A journals.
Elected President, Graduate Common Room, Durham University, 2013 to 2014.
Elected International Students Officer, Graduate Common Room, Durham University, 2011 to 2013.
Volunteer work with the Ustinov Seminar, the Durham and Mauritius Rotary Club, the Durham University Chinese Students and Scholars Association, and Ustinov Volunteer Opportunities.
More on the Research page.
The work below rests on four positions about evidence. They are what separate a defensible finding from a well-fitted one.
Regression, machine learning and data mining are all correlation without an identification strategy, and adding controls does not change that. A causal claim needs a design, an identification strategy, and tests built to break the result rather than confirm it.
Policy changes and other external shocks create natural experiments already sitting in the data: treatment and comparison groups formed by something that happened anyway. Finding them and analysing them is cheaper than any trial and often closer to the real world. I have run these for clients.
Not only a research interest: it is how the platforms are designed. The finite intelligence budget in AXIOM, the real-data economy in Miniconomy and load-bearing knowledge in MoneyWorld are experimental designs, not just mechanics.
Survival analysis combined with neural approaches, and similar pairings: the rigour of econometrics with the flexibility of machine learning, for accuracy without losing the ability to defend a result.
My research is in behavioural economics and behavioural finance: how people actually make financial decisions, and what those decisions do to them. It runs across human decision making, financial capability, financial wellbeing, emotional wellbeing and overall wellbeing. None of it was ever research for its own sake. It is how I came to understand people, organisations and systems well enough to build for them, and it is the foundation every platform on this site is built on.
Psychological Distress and Household Net Worth
Households facing persistent psychological distress accumulate less wealth across the life cycle, and the gap survives standard income, employment, and demographic controls. We ask whether mental-health stress operates as a measurable behavioural-economic friction on household balance sheets, distinct from income shocks and education channels.
Online IDs: DOI · RG 362161202
Financial Literacy, Trust, and Stock Market Participation
Stock-market participation rates differ sharply across households with comparable income and wealth, and the residual gap maps onto differences in financial literacy and trust. We ask whether financial literacy and trust in financial institutions operate as behavioural-economic prerequisites for market participation, distinct from access and cost channels.
Online IDs: DOI · RG 264786037
PhD Thesis: Durham University Business School (2015)
Doctoral thesis comprising empirical essays in household finance. Applies modern panel-econometric techniques to large household panel datasets to investigate financial decision-making, wealth accumulation, stock market participation, trust, and the behavioural determinants of household financial outcomes.
Funder: Durham University Business School (doctoral funding) · etheses.durham.ac.uk
Academic conferences and invited seminars.
Melbourne, Australia
Dynamics of Household Financial Hardship: Landing, Path, Duration and Exit. Pho, L.T.; Qiu, M.; Balloch, A. Presented in the behavioural finance stream.
Melbourne Institute of Applied Economic and Social Research, University of Melbourne
Dynamics of Household Financial Hardship: Landing, Path, Duration and Exit. Pho, L.T.; Qiu, M.; Balloch, A. Presented at the conference run by the Melbourne Institute, which produces the national household panel survey the paper is built on.
Massey University, New Zealand
Miscellaneous seminars, workshops and guest presentations across schools and programmes.
Te Ara Ahunga Ora Retirement Commission: internal seminar presentation
Government vs Community Support in Times of Crisis: when someone falls into financial hardship, are they helped more by government support or by their family and friends?
Held remotely, USA
When It Rains It Drains: Psychological Distress and Household Net Worth. Balloch, A.; Engels, C.; Philip, D.
Auckland, New Zealand
Self-efficacy and poverty transitions, using data from the Panel Study of Income Dynamics (PSID).
Manchester, UK
Financial distress, poverty and financial decision making. Balloch, A.; Philip, D.; Nicolae, A.
Durham, UK
Household financial literacy and financial management.
Karachi, Pakistan
Autonomous Helicopter: Model, Simulation and Control.
This is how I work out what to build and how to build it. The questions here come from the same places my platforms do, household finance, wellbeing and the decisions people actually face, and what I learn feeds the products, the industry analytics and the policy work rather than a publication count.
Economic Shock, Hardship Landing, and Intra-Household Financial Anxiety Contagion
Telework and Sole-Parent Financial Wellbeing
Self-Efficacy, Wealth Formation, and Household Resilience
Locus of Control and the Persistence of Financial Hardship
Telework, Lockdowns, and Anxiety Across 151 Countries
Family Financial Socialization and the Making of Capability
Financial Parenting, Capability, and Adult Wellbeing
Empirical Evidence from New Zealand's Labour Force during COVID-19
Learning Disability, Financial Hardship, and Recovery Dynamics
Work-Family Attitudes, Sole Parenthood, and Hardship
Financial Decision-Making, Relationship Strain, and Separation Risk
Women's Financial Capability and Recovery After Life Events
Gender, Financial Risk-Taking, and Hardship Recovery
Third-Tier Lending, Housing Stress, and Financial Vulnerability
Social Engagement, Welfare Regimes, and Financial Vulnerability
Disability, Debt Management, and Retirement Preparedness
Why Model Choice Matters in Targeting Australian Financial Hardship
Latent Classes in a 7-item Hardship Module and What They Predict for Subsequent Recovery
How Where You Move From Shapes What Happens to You: A Multi-Shock Panel Analysis
A Competing-Risks Mediation Analysis
Culture, Trust, and Financial Advice-Seeking in Aotearoa
Saving Behaviours, Debt Management, and Future Financial Security among Individuals with Disabilities
A Machine Learning and Behavioural Perspective
Policy Reforms, Economic Shocks, and Household Resilience
My own research using the Stats NZ Integrated Data Infrastructure (IDI): de-identified, linked government administrative and survey data, accessed under Stats NZ’s approved-researcher rules.
