Back to selected work

Résumé

William Jeffries

AI Product Engineer

william@williamatwork.com williamcodes williamjeffries

Profile

Product-minded engineer with 12+ years of experience building production systems across healthcare, banking, fintech, media, education, and public-interest technology. I take AI products from domain discovery through models, evaluation, infrastructure, and operations.

Experience

2025–present

Independent AI Product Engineering

AI Product Engineer

  • Built a multilingual research agent over 87,218 historical page scans, including vision transcription, diacritic-aware full-text search, text and visual embeddings, 11 agent tools, and a bilingual research interface.
  • Built production workflows for roughly 10 learning-asset types across a 191-course catalog, with LangGraph evaluators, restartable operators, and browser automation for systems without APIs.
2020–2025

Parachute Health

Senior Software Engineer → Head of DevOps → AI Architect · Contract

  • Led the core AI service for a HIPAA-regulated medical-document intake product, from architecture and model selection through AWS infrastructure, stakeholder work, and deployment.
  • Built offline hallucination and accuracy evaluators against golden datasets and a LangGraph agent that generated and validated the company’s proprietary configuration DSL.
  • Led cloud and delivery infrastructure across Kubernetes, Terraform, Helm, Argo CD, and Buildkite during a five-year embedded engagement.
2016–2020

Stride, Contino, and independent engagements

Senior Software Engineer & Consultant

  • Led product and platform work for clients including IBM, JPMC, JOOR, HBC, and TD Ameritrade across application development, cloud infrastructure, SRE, and engineering practice.
  • Led a team of 10 at JOOR and helped open its European engineering office; delivered systems serving up to 500,000 users and 3,000 requests per second at IBM.
2014–2016

Heat Seek NYC and SunGard

Co-founder / CTO · Senior Developer / Consultant

  • Co-founded Heat Seek, building web-connected temperature sensors and software around the workflows of tenants, organizers, and legal-aid lawyers. The organization and its partners served more than 650 New York City families.
  • Built a Polymer application used by more than one million Comcast customers at SunGard and was promoted within six months.

Selected earlier work

Accent Trainer · Founder. Designed a PyTorch accent-classification model and led its production deployment on SageMaker.

Pixelgrid · Founder. Trained a TensorFlow computer-vision model for accessibility analysis inside a self-built Kubernetes platform.

Education & public work

College of William & Mary

RailsConf speaker, open-source maintainer, and former producer and co-host of The Rabbit Hole developer podcast.