Winsley Jeune · Applied AI Engineer

Applied AI systems. Built in public.

After years of shipping production software across data and full-stack engineering, I'm now building applied AI systems focused on extraction, retrieval, automation, and evaluation.

Every system is public — code, architecture decisions, benchmarks, failure reports, and post-mortems included.

Winsley Jeune

Current focus

Applied AI, end-to-end, using the primitives directly. Three areas:

  • Structured extraction — LLM tool-use with strict schemas, replacing brittle scrapers.
  • Retrieval-augmented generation — chunking, embeddings, pgvector, and citations that hold up.
  • Agents and evals — small, tool-using loops with regression tests, not demoware.

Selected projects

All projects →

Quantara — product extraction backend

Live

Self-hosted scraper for e-commerce research. Paste a product URL or PDF catalog; Claude returns a typed JSON product (title, SKU, price, specs, variants, images) and joins a batch that exports to multi-sheet Excel. No per-vendor selectors — text path with vision fallback.

  • Claude Sonnet 4.6
  • Puppeteer
  • Node.js
  • SQLite

Regulatory intelligence RAG platform

Currently shipping

Retrieval pipeline over public regulatory documents: ingestion, semantic chunking, embeddings, pgvector indexing, and a Next.js research interface with source-grounded citations. Includes golden datasets, regression testing, and retrieval benchmarks to prevent prompt drift.

  • RAG
  • Embeddings
  • pgvector
  • Next.js
  • Evals

Notes on shipping applied AI — what works, what breaks, and what I changed after the second time it broke.

Get new posts in your inbox

Notes on shipping applied AI: scrapers, RAG, agents, evals. No fluff.

One email per post. Unsubscribe anytime.

Background

Currently Software Engineer at Suncor Stainless in Plymouth, Massachusetts. Eight years across data analysis, full-stack engineering, and applied AI. Open to senior and staff applied-AI roles — Boston, hybrid, or remote ET.

Hiring an applied AI engineer?

The fastest way to reach me is email. Résumé and links below.