Work

Systems we have built and run.

Client names are kept private. Everything else is described in full: the problem, the architecture, the integrations, the security work and the outcome, including where a project is still a pilot.

Operations · Finance

Racing Operations Platform with Per-Horse Profit and Loss

A multi-site thoroughbred training business had its money in Xero, its horses in a stable-management product, its staff in an HR platform and no way to see per-horse profitability. Now it has one reconciled view, on the web and on the phone.

PythonFastAPIPostgreSQL 16SQLAlchemyAlembic
Read the case study →
Operations · Workforce

Stable Scheduling Platform Replacing a 52-Tab Spreadsheet

A 545-horse, three-site stable ran its daily training from a 52-tab Google Sheet with no validation, access control or audit trail. It now runs from a rules-checked scheduling platform with TV boards, reports and a rider app.

PythonFastAPIJinja2PostgreSQLSQLAlchemy
Read the case study →
Construction · Legal

Building Compliance Knowledge Base for Building Surveyors

A building surveying firm needed every clause of the Building Act, Regulations, NCC and regulator guidance searchable with a citation. The core corpus of 463 documents and 45,000 clauses went live in about two days.

PythonFastAPIPostgreSQL 18Full-text + trigram searchServer-rendered HTML
Read the case study →
Sport · Education

Rowing Coaching Hub for a School Program

Sixteen coaches, forty boats, a weekly roster in a spreadsheet and the week's plan in WhatsApp. Now one installable hub with rosters, session plans, a 300-drill library, crew boards, erg results and live lake conditions.

Django 6PostgreSQLRenderProgressive web appWeb Push
Read the case study →
Data platform

Horse Racing Database with Public API and AI System Builder

A dozen sources, seventeen scheduled jobs, 489,000 races and 2.2 million results reconciled into one database with a fast web front end, a read-only API, a native iOS app and an AI query tool that backtests its own answers.

Python 3.11FastAPISQLAlchemy 2PostgreSQLDuckDB
Read the case study →
Computer vision · Research

Equine Video Biomechanics: Measuring Yearlings and Breeze-Ups from Sale Video

A custom 45-point pose model applied to 37,800 yearling parade and breeze-up videos, with repeatability gates, measurement of stride, reach and conformation, and validation against millions of race results. Live on public horse pages.

PythonOpenCVYOLOSAMDeepLabCut
Read the case study →
Computer vision · IoT pilot

Daily Barn Gait Monitoring with Raspberry Pi Cameras and Serverless Vision

Two barn cameras, a Raspberry Pi, cloud storage, serverless scanning, GPU pose estimation and vision-model brand reading, built as a modular pipeline to compare each horse's walk with its own history. A working pilot, honestly reported.

Raspberry PiIP cameras (RTSP)Google Cloud StoragePub/SubModal
Read the case study →
Consumer apps · ML

Racing Ratings Engine, Daily Picks Apps and a Public Scoreboard

A weekly-retrained ranking model, an explainable scorer that writes its own reasoning lines, a resilient price feed, consumer apps and a public scoreboard that shows the losses as well as the wins.

PythonLightGBMFastAPIPostgreSQLRender
Read the case study →
Media automation

Automated Video Production from a JSON Spec

Give it a JSON spec and it returns a finished 1080p video: presenter clips, themed cards, AI narration fitted to a time cap, music, sound effects, subtitles and an end card.

PythonffmpegElevenLabsHeyGenPillow
Read the case study →
Agribusiness · In-house In development

Agribusiness Operations and Livestock Intelligence Platform

An in-house platform now being scoped for livestock and mixed farming enterprises: paddock and mob ledgers, per-mob profit and loss reconciled to Xero, weigh and NLIS data, camera and sensor monitoring, and an offline paddock app. Built on patterns proven in racing.

PythonFastAPIPostgreSQLXero APINLIS data
Read the case study →

How to read these

Each case study says what was built, what it runs on and what it costs to host. Where a system is live in daily use we say so. Where it is a research pilot, we say that too, and we are specific about what has and has not been validated. We would rather you trust the modest claims than doubt the big ones.

Have something you need built?

Tell us the problem, the systems you already use and what a good outcome looks like. You will get a plain-language reply within one business day, usually with a first sketch of how we would build it.