AI Field Engineering · Product · 15 years
I take AI from the
boardroom to production.
I lead AI agent deployments end to end — owning the executive relationship and the technical delivery on the same engagement. Then I go build the thing: agentic infrastructure, production ML systems, evaluation harnesses. Strategy and shipped code, from one person.
What I actually do
the rare combinationLead the deployment
I own the strategic layer between a customer's business problem and the AI solution — solutioning, executive relationships, go-live readiness. I've shipped AI agents into the highest-stakes enterprise environments there are.
Build the system
I don't hand off to engineering and hope. I build the agentic infrastructure and ML systems myself — MCP servers, retrieval pipelines, production models behind APIs, evaluation harnesses treated as first-class.
Translate both ways
Boardroom to codebase and back. I can size a deal with a CRO in the morning and debug a serving pipeline in the afternoon — the rare combination that turns AI pilots into production revenue.
Selected Work
built, not describedMLflow MCP Server
open sourceAgentic infrastructure: a Model Context Protocol server that gives Claude direct read/write access to MLflow experiment tracking — six tools spanning run search, logging, and tagging. Dual transport (stdio + SSE), Docker-ready, stateless by design. The connective tissue AI-native teams are racing to build.
AI Customer Retention Platform
liveChurn prediction on KKBox's 31GB dataset — 12 models benchmarked, LightGBM champion at 0.966 AUC. The real work is the decision policy: I showed the naive 'target the top 10K riskiest' approach loses money (−$6.8K), then tuned a cost-sensitive threshold that turns the same model ROI-positive. Shipped with a live scoring demo, batch upload, and an ROI simulator.
FinSight — Financial RAG
in progressAgentic retrieval over four financial modalities — SEC filings, earnings transcripts, OHLCV market data, and news. LangGraph-orchestrated, hybrid dense + BM25 in one Qdrant index, Cohere reranking, RAGAS evaluation. The differentiator is evidence-conflict detection: flagging when sources disagree rather than silently picking one. Design targets: faithfulness ≥ 0.80, P95 ≤ 3s, ≤ $0.005 / query.
Track Record
15 years · MS CSAI Deployment Strategist
SalesforceLead AI agent (Agentforce) deployments for the most demanding enterprise customers — owning executive relationships, solutioning, and go-live.
Technical Architect → Sr. Architect → People Manager
SalesforceProgressed through senior technical and leadership roles across high-tech, public sector, financial services, and healthcare accounts.
Senior Consultant
DeloitteEnterprise technology consulting — the foundation of fifteen years translating between business strategy and technical delivery.
M.S. Computer Science
University at Buffalo, SUNY
For fifteen years I've sat where business strategy meets technical delivery. Today I lead enterprise AI agent deployments — owning executive relationships, solutioning, and go-live for some of the most demanding customers in the market.
What sets the work apart is that I don't stop at the slide. I build the systems underneath — agentic infrastructure like MCP servers, production ML pipelines with real experiment tracking, and evaluation harnesses I treat as first-class. That dual altitude — fluent with a CRO and fluent in the codebase — is how AI pilots actually become production revenue, and it's what I bring to a field-engineering or AI product leadership team.