RAGEvalsData ContractsProduction
RAG in the real world: evals, drift, and data contracts
Building RAG systems that hold up in production requires more than good retrieval. Here's what actually breaks — and how to catch it early.
·5 min read
Ideas, tested
Research
Our own studies in fine-tuning, context engineering and graph engineering. We publish what we learn, including what didn't work.
This isn't a service. It's how we stay good at the two things that are.
Notes
Entity ResolutionFinanceLLMsData Engineering
Entity resolution for financial instruments: deterministic + probabilistic + LLM assist
Financial data is full of the same entity described dozens of different ways. Here's a layered approach to resolving them reliably.
·6 min read
AgentsOrchestrationObservabilityLLMs
Agent workflows that don't break: orchestration, guardrails, and observability
Multi-agent systems are powerful and brittle. Here's the engineering discipline required to make them reliable in production.
·7 min read
Research projects
Studies run as projects
Studies run as projects rather than write-ups. They appear in the showcase under the AI research Pillar.
Nothing in the showcase under this Pillar yet.
The notes above are where this work is published today. Every Project
Your next move