Popular repositories Loading

1. amazon-ppc-evals amazon-ppc-evals Public

   Evaluation suite for an LLM search-term classifier (Promptfoo, Claude + GPT): human-labelled cases, a scoring rubric, and a documented root-cause fix that lifted judge pass rate from 60% to 90%.

2. staffflow staffflow Public

   AI candidate-matching PoC for staffing agencies — Supabase + Anthropic with a validation gate. A Solutions Architect work sample.

   TypeScript

3. inspectiq inspectiq Public

   Document-intelligence + invoice anomaly pipeline over PDFs (Qdrant, multilingual e5, Claude) with a human-approval validation gate

   Python

4. evalops evalops Public

   CI-gated evaluation + observability for LLM/RAG systems: a regression gate that fails the build when retrieval/answer quality drops below a committed baseline, with per-query observability traces.

   Python

5. connectai connectai Public

   Production-grade RAG-driven customer-support agent for an international telecom service — hybrid retrieval, reranking, grounded generation with an abstention gate, CI-gated eval, and observability.

   Python

6. traceforge traceforge Public

   Requirements traceability & compliance copilot for safety-critical systems — grounded RAG with classification-aware access control, medallion lakehouse, MCP server, CI-gated eval. Built on Azure/Py…

   Python