Selected work
Selected work for AI Systems
Prioritized and described for a ai systems audience — switch lenses to re-rank them. Most were built in client and production environments, so the source is proprietary; I'm always happy to walk through the architecture and trade-offs.
AgentZero
Autonomous AI coding agent
A local-first autonomous coding agent built on a multi-agent state machine spanning planning, execution and validation. AST-aware semantic retrieval (Tree-sitter + vector DB) gives high-precision code context at the symbol level, and a self-healing execution loop runs inside a secure Docker sandbox with adaptive model routing for reliability and cost.
symbol-levelcode retrievalself-healingexecution loopsandboxedDocker isolation
PythonMulti-AgentTree-sitterVector SearchDockerLLM Routing
Conversational Analytics Platform
Natural-language querying over Snowflake
A conversational analytics platform for natural-language interaction over multi-domain marketing & retail datasets on Snowflake. Strategy-specific LLM agents sit over semantic views (UMM, MTA, Sonic, Retail, EDM) with intent routing and dynamic tool selection, turning simulation and optimization workflows into conversational execution — deployed via Streamlit on Snowflake.
5strategy agentsmulti-domainNL querying
Snowflake CortexCortex AgentsStreamlitLLM RoutingPrompt Engineering
AI-Driven Financial Data Platform
Production RAG for real-time financial insight
A production RAG system for real-time financial insight — ingestion, vector indexing, hybrid retrieval and LLM-orchestration APIs. Tuned embedding retrieval and batching to cut model-query latency by ~25%.
−25%query latencyhybridsearch retrieval
FastAPIPostgreSQLRAG PipelinesVector DBsReact
Enterprise Event Logistics Platform
Microservices for large-scale event operations
A microservice backend for large-scale event logistics with an optional natural-language interface — LLM APIs let operators query schedules conversationally over independently scalable booking and user services.
0downtime deploys−30%debugging time
FastAPIPostgreSQLMicroservicesDockerAWS