- Per-layer
- contract enforcement
The problem
Data contracts break silently. A producer changes a field, and the consumer three services downstream fails hours later with an error that points nowhere near the cause.
Approach
Make the contract graph a first-class artifact — render every producer/consumer edge and validate changes against it before they ship.
Architecture
- Service and schema definitions are parsed into a directed graph of producers, consumers, and the contracts between them.
- Policy rules evaluate proposed schema changes against downstream consumers and flag breaking edges.
- Interactive visualization highlights violations in place rather than in a log.
Design decisions
Visualize the graph rather than emit a report
Contract violations are inherently topological — the question is always "who downstream breaks". A list of errors loses the structure that makes the answer obvious.
What was hard
- Graph layouts become unreadable past a few dozen nodes, so the view had to collapse to the affected subgraph rather than render everything.
What I took from it
- Enforcement only works where it is cheap to obey. Surfacing violations at change time instead of at runtime is the whole value.
Stack
Other work
- Product
Atriveo
Job-search platform with a 5.0-rated Chrome extension and live customers.
- Applied Research
FOMC Intelligence
NLP pipeline turning Federal Reserve communications into market signals.
- AI Systems
Legal RAG
Retrieval-augmented generation over legal filings, with citations that hold up.