AI Product Engineer · Production Agents · Complex Domains
Proof in production
Grew Aligrade's product to $300K in revenue with two top-10 pharmaceutical customers after it acquired Grantease's technology. Previously scaled Grantease to $100K ARR in three months, with adoption at 30+ universities—including Stanford, Harvard, UPenn, Duke, UCLA, Mount Sinai, and Scripps.
Built production agent systems end to end—Pydantic AI supervisor/subagent orchestration, hybrid retrieval and parallel LLM reranking, validated outputs, collaborative document editing, resilient model routing, and agent-driven end-to-end evals.
Turned ambiguous expert workflows into production AI products—from direct user discovery and product definition through architecture, evals, full-stack delivery, and production reliability.
As Lead AI Engineer at Aligrade, owned AI architecture, technical roadmap, hiring, sprints, and delivery while leading three engineers and a product manager—and personally building the core systems. Published first-author AI evaluation research with Springer, co-authored LLM benchmarking with Lenovo and AMD, and wrote a three-part enterprise AI engineering series.
Beyond the work
I am from Philadelphia, PA. I studied at Duke University. I currently live in Denver, CO.
I like playing and designing strategy board games, running, and tennis. My most useless skill is Etch A Sketch.
Selected work
I joined after Aligrade acquired Grantease's technology. I led the team building production AI agents, owning AI architecture, the technical roadmap, engineering and product hiring, customer discovery, and delivery. The product grew to $300K in revenue with two top-10 pharmaceutical customers.
A Pydantic AI multi-agent system with supervisor/subagent orchestration and validated outputs for regulated workflows; an AI-native collaborative document editor for precise, reviewable edits; hybrid BM25 and semantic retrieval with parallel LLM reranking; and a CLI that made the production application fully operable by agents for deeply interactive end-to-end evals.
I bootstrapped and scaled Grantease to $100K ARR in three months. The platform produced 1,000+ NIH proposals for researchers at 30+ universities before its technology was acquired by Aligrade, where I joined as Lead AI Engineer.
A Next.js collaborative editor with an AI writing agent, Supabase and WorkOS enterprise authentication, DOCX/PDF ingestion, resilient fallback and structured-output recovery across five LLM providers, and an academic citation engine with automated data cleaning and Crossref metadata enrichment.
I built and shipped RAG, NLP, and transformer products for three Fortune 500 customers, including enterprise knowledge retrieval, executive competitive intelligence, automated feedback routing, and assistive writing for people with ALS.
Customer-facing RAG systems, evaluation pipelines, data workflows, low-latency transformer inference, and production deployment for on-premises enterprise environments.
Editor MCP Public repository · Demo-ready
I built a TypeScript MCP server and Tiptap/Yjs editor where agents make atomic, reviewable document changes instead of pasting text into chat. People can inspect, navigate to, accept, reject, or rewrite every proposed edit.
View source and demoVersioned semantic tools with revision guards and atomic edit groups
Live shared state through Tiptap, ProseMirror, Yjs, and Hocuspocus
Authenticated Streamable HTTP with OAuth and fail-closed validation
First author · TPCTC 2024
Designed and ran an experimental pipeline that generated synthetic question-answer datasets with four open-source LLM configurations, fine-tuned the same base model on each, and tested which dataset signals predicted downstream QA performance.
Found that semantic similarity can indicate domain diversity and that short chain-of-thought answers may flag lower-quality training data.
Co-author · TPCTC 2023
Co-authored a Lenovo and AMD review of how to compare LLM systems across training and inference—not only on accuracy and throughput, but also cost, energy, bias, trust, and sustainability.
Published by Springer in Lecture Notes in Computer Science; the chapter has been cited in subsequent benchmarking work.
Lenovo Press · Three-part engineering series
I co-authored a practical series for enterprise teams building domain-specific RAG systems, from model selection and evaluation through synthetic dataset creation, LoRA fine-tuning, hardware sizing, latency, and production tradeoffs. I also developed a containerized framework that customers could use to implement the approach.
Graduate AI coursework · 2024
Designed a card-shuffle simulator and trained encoder-only PyTorch transformers to test how model depth and increasing randomness affect ace-sequence prediction. Deeper models learned subtle structure in low-noise sequences; performance degraded as shuffle complexity increased.
Graduate AI coursework · 2024
Collected gameplay trajectories and trained PyTorch policies with behavioral cloning and DAgger for two-agent control. The strongest independent-policy pair scored 18 goals across 32 evaluation games and held the same scoring rate against unseen agents.
Contact