FDE AI | New York
Client-facing AI engineer building production agentic systems end-to-end, from C-level discovery to architecture and deployment. Led a $2M AI transformation for $90B AUM PE firm with a team of 8. Deployed 12+ GenAI systems with production architecture, evals, governance, and observability. 2X AI and robotics founder (Techstars, NVIDIA Inception).
| Outcome | Where |
|---|---|
| Due diligence 4 months → ~1 hour | $2M program, $90B AUM PE firm |
| 93% Recall@10 on a 100+ question golden set | Regression-gated enterprise RAG-KAG |
| Technology adoption 6 weeks → 3 | 5-agent platform, 750 documents |
| Integration effort −60% | MCP server over 18 internal APIs |
| Data retrieval hours → minutes | GraphRAG copilot, ~3,000-entity graph |
| ~15 h/week returned to supervisors | Multimodal vision AI, 5 VLM agents |
Discovery with the people who own the P&L. Scope to one workflow with a measured baseline. Ship the thin production path first: retrieval, evals, guardrails, tracing. Prove it against the baseline and publish the misses. Hand off a runbook, eval gates, and owners.
Regulated, NDA-bound environments. Client code stays private, so the repos below are open rebuilds of the same patterns on public data.
| Project | What it is | Status |
|---|---|---|
| Rivet-KAG | Hybrid KAG over your documents, answers cited to source | 🟢 In progress |
| AgentGate | Auth, routing, and audit for every agent and MCP call | 🟡 Design |
| EvalForge | Auto-generated golden sets and red-team evals | ⚪ Planned |
| OpenDiligence | Multi-agent diligence over SEC EDGAR filings | ⚪ Planned |
| RegGraph | Regulatory change mapped onto policy and controls | ⚪ Planned |
Each ships with an architecture diagram, ADRs, an eval scorecard with real numbers, and cost/latency per query. If a number is bad, it goes in the README anyway.
Python · TypeScript · LangGraph · LangChain · LangSmith · MCP · GraphRAG · ontologies · Neo4j · Milvus · FastAPI · PostgreSQL · Redis · Docker · Azure · AWS · GCP · Palantir AIP