Agentic workflows, hybrid retrieval over graph and vector stores, and quant research platforms cited in NeurIPS and ICML.
An AI-powered financial risk intelligence agent built with LangGraph and Claude Sonnet 4, connected to 163 tools via the SAJHA MCP Server. Uses SSE streaming for real-time analysis.
A retrieval-augmented generation agent answering questions about OSFI's Capital Adequacy Requirements — Canada's primary banking capital regulation. Achieved a 0.805/1.0 quality score.
A hybrid AI search and retrieval system combining a Neo4j graph database with Pinecone vector search, deployed on AWS. LangChain orchestrates queries needing both semantic understanding and relationship traversal.
An AI-oriented quant research platform covering the full pipeline from data ingestion to live trading, with an autonomous RD-Agent LLM that generates and tests strategies. Cited in NeurIPS and ICML papers.
Multi-tool agents with orchestration and guardrails.
Graph + vector retrieval for queries that need both.
ML platforms rigorous enough to be published.
Eval-driven · observable · production-grade AI. Tell us about your problem and we'll tell you honestly how we'd solve it.