Applied AI Product Engineer
Product-minded AI engineer · Search, RAG, recommendations, learning systems
I’m a software engineer with 12+ years across full-stack product engineering, platform work, and applied AI. Currently building AI systems for medical education: semantic search, RAG, recommendations, and personalized learning workflows for clinicians.
Washington-based · Hawaii/HST-friendly · U.S. remote
Work I’m drawn to
I like the messy middle between product, engineering, and real users: understanding the workflow, building the first useful version, integrating with real systems, measuring whether it works, and turning repeated needs into reusable product capabilities.
This overlaps with work often called forward-deployed engineering, AI product engineering, applied AI engineering, or product engineering for AI systems.
Recent applied-AI work for medical education, used in production by thousands of clinicians.
Retrieval-augmented search
Built retrieval-augmented semantic search to help clinicians find relevant learning content by meaning, not just keywords. Work included embeddings, pgvector/Postgres, retrieval strategy, product integration, and production rollout.
Personalized recommendations
Built personalized recommendations that adapt to member behavior and content relationships. Designed the recommendation logic, data model, retrieval strategy, and full-stack product integration.
Workflow AI
Turned group-level assessment misses into targeted study recommendations. Connected assessment data, content mapping, retrieval, reranking, and product UX to help teams identify and close knowledge gaps.
In progress In progress
Building a daily learning plan that combines spaced repetition, recommendations, and retrieval to help clinicians retain important material over time.
Featured case study
A look at recent applied-AI work in medical education: semantic search, RAG, recommendations, and personalized learning workflows designed to help clinicians find and retain relevant material.
Read the case study ↗I’m comfortable starting with unclear requirements, legacy constraints, partial data, and real users who do not care about AI for its own sake. The work I enjoy most is turning that ambiguity into useful, reliable product behavior.
Understand the workflow, user pain, constraints, and success criteria.
Prototype quickly, then turn the useful parts into production-grade systems.
Connect AI features to existing data models, APIs, auth, product surfaces, and operational workflows.
Evaluate relevance, adoption, latency, cost, reliability, and user trust.
Turn repeated needs into reusable product capabilities instead of one-off hacks.
Tools I reach for
AI / retrieval
OpenAI, embeddings, RAG, reranking, pgvector, vector search, LangChain, evaluation patterns
Product engineering
TypeScript, React, Node.js, Postgres, GraphQL/REST, full-stack product architecture
Production reality
APIs, data modeling, observability, auth, performance, cost/latency tradeoffs, production debugging
The record
Designing and shipping production AI product systems for medical education, including semantic search, RAG, recommendations, assessment-based study plans, and personalized learning workflows for clinicians.
Led and built core systems for sales, fulfillment, and operational workflows, including event-driven services, internal tooling, and platform features.
Built checkout, payment, enrollment, and internal sales tools for a high-throughput education marketplace.
Let’s talk
I’m always interested in thoughtful conversations with teams working on applied AI, learning systems, search, recommendations, workflow automation, or production LLM systems.
p@patrickeddy.comWashington-based · Hawaii/HST-friendly · U.S. remote