Projects
Employee recognition, retail, gaming, venture capital, aviation training, procurement, maintenance, loyalty, logistics, insurtech, building materials and professional golf. One pattern throughout: study the business first, then build the system it actually needs.
Warehouses, pipelines and the BI layer on top — owned by the client, not rented from a vendor.
Sisense replaced by a customised Superset the client owns outright, on rebuilt pipelines and a purpose-built analytical warehouse — embedded back into their own product, at CAD 600k below the licensed alternative in year one.
$600k
CAD below the licensed alternative, year one
A complete BI ecosystem the client owns outright, built instead of signing for an enterprise SaaS platform priced to grow with them.
$200k+
saved annually, recurring
Ingestion, a layered warehouse and an ML workspace over a live game economy — with a conversational layer that queries Snowflake directly through MCP.
0
SQL required to ask a question
Agents that do real work against real data, with gates that hold when the model is wrong.
A digital analyst that runs the top of the funnel autonomously: hundreds of millions of rows a week, a custom multi-agent router, and one-button outreach written back to the CRM.
300%+
faster deal review
We paused the BI work we were paid for to migrate the platform first — then built the KPI warehouse and shipped Eddy, an agentic assistant deployed as a first-class Open edX XBlock.
5–6×
increase in company valuation
A supervised agent system with its limits designed in rather than documented: Temporal owns the workflow, and a deterministic policy engine can overrule the model.
Fail-safe
deterministic gate over model output
A quoting system grounded in the client’s own casebook and historical jobs, built alongside structured discovery that documented what the business needed it to decide.
Own
casebook, not generic knowledge
Full systems built, deployed and kept running — often as the client’s entire technology department.
Three coordinated workstreams as their embedded engineering team: a live Java ERP kept shipping, a headless storefront built from scratch, and a migration toolkit that moved the data between them.
5 weeks
to launch readiness
They did not need a tool; they needed a technology department. Two years on we are still shipping — and an AI agent now ranks technicians for each event.
75%
projected cut in booking headcount
A cartographic application for insurance agents, inherited with more than fifty critical bugs and a budget nearly gone. We fixed the structural causes and deployed it — because the previous team never had.
50+
critical bugs cleared to a shipping MVP
Years of Excel macros and MS Access reverse-engineered into the first complete blueprint of the business anyone had produced — then a phased migration, and a compromised estate rebuilt and secured.
1st
complete blueprint of their processes
Every shot a player used to write down by hand — club, lie, number on the card — captured as structured data, modelled into hole-by-hole game plans, and updated live against tournament feeds while the round is still being played.
Live
tournament feeds driving in-round strategy
Your project
Bring us the problem in whatever shape it is currently in — a spreadsheet that has outgrown itself, a stalled build, a data set nobody can query, an AI idea you are not sure is feasible.
