Cases
Every case below is a system that runs today — not a pitch deck. We show the challenge as the client framed it, what we built, and the result we can stand behind. Where a number comes from someone else, we say so.
Market intelligence · Agri-commodities
Traders, processors, and exporters in the global grain and oilseed markets buy and sell against prices that move every day. Getting those prices late — or getting them wrong — is expensive.
Our role: developer on a long-running engagement — a decade of building and maintaining the platform, not a one-off delivery. Built on Django, PostgreSQL, and Celery.
Tender & RFP analysis · Ireland
A bid team can only chase so many tenders at once. The cost of qualifying one is measured in senior people's days — and most of that time is spent on tenders they end up not bidding.
Client confidential at their request. We can discuss the outcomes on a call; the internals stay private.
Financial-statement extraction · SEC filings
Analysts who compare companies need the numbers in rows and columns. The filings that contain those numbers are written to be read, not to be parsed.
Delivered as a document-processing pipeline; accuracy measured against manually verified filings.
From our R&D
Some capabilities we develop on our own R&D projects first, so the method is tested before a client pays for it. These two are R&D we ran — not client production systems — and the numbers below come from those R&D runs.
We migrated a 44,000-line PHP codebase to Python unit by unit — 11 of the 14 units, 79% of the system, with 918+ tests written along the way and zero broken API endpoints. The point of the exercise: prove a legacy system can move incrementally, staying usable the whole time, instead of going dark for a rewrite.
Read the full case →We built a gateway that sits in front of AWS Bedrock and enforces budgets per team, with full observability over what is being spent and by whom. Across the R&D budgets we ran through it, the overshoot stayed under 1% — the difference between AI costs you can plan around and AI costs you discover at the end of the month.
Read the full case →Tell us the one document type that eats the most of your team's time. We'll come back with what AI can realistically do with it — on your data, with the numbers we'd be willing to publish.
