About
Software engineer who sits with teams, finds the bottleneck, and ships production AI workflows: agents, extraction, and retrieval. Owns reliability after ship: evals, cost, latency, and failure modes. Python and TypeScript. Founder of Mellow AI. Author of LangChain's Salesforce integration. Contributor to LiteLLM (50+ PRs) and Pydantic AI.
Experience
Founder — Mellow AI
2025 – Present- •Embed with teams to turn fuzzy AI problems into scoped, shipped systems spanning extraction, agents, and evaluation
- •Built openextract, a model-agnostic document extraction toolkit (Python, Pydantic) used in production workflows
- •Owned cost and latency tradeoffs on custom OCR: 95% cheaper and 75% faster than the prior path
- •Upstreamed production patterns into LiteLLM (50+ merged PRs) and Pydantic AI
AI / Data Engineer — Pay Ready
2023 – 2025- •Shipped production AI agent workflows that cut manual effort 60%, then owned them after the first version
- •Built a multimodal document and text pipeline (+25% accuracy) and debugged model failures in real traffic
- •Shipped Slack RAG bots used in daily work, cutting retrieval time 75%
Scaled Tooling Analyst — Otter / CloudKitchens (now Atoms)
2022 – 2023- •Embedded with ops and engineering to ship internal tooling: Python pipelines scoring ~500GB/week
- •Shipped a Jira chatbot from discovery through production. Resolution times improved 30%
Business Analyst — RealPage
2021 – 2022- •Partnered with 75+ internal users to ship SQL models and dashboards
- •Automated reporting pipelines, cutting manual workload 25%
Open Source
LangChain-Salesforce — Author of the official Salesforce integration
Pydantic AI — Contributor (xAI tools and model settings)
LiteLLM — 50+ merged PRs on the LLM gateway
SmolAgents — Contributor to Hugging Face's lightweight agent framework
Projects
Open Lease Audit — Streaming lease extraction and cross-portfolio integrity console (Next.js, TypeScript). Schema, extract, audit, with evidence and confidence per field.
openextract — Open-source document extraction toolkit (Python, Pydantic). Turns PDFs, images, and text into structured data using LLMs. Type-safe, model-agnostic, 1K+ PyPI downloads.
pdfmd — Zero-dependency Rust PDF to Markdown converter. About 4ms on a 17-page paper. Owned the parser, decoder, and heuristics end to end.
Education & Certifications
- •RAG, LangChain, LangGraph, Semantic Search, DeepLearning.AI
- •Azure Applied AI, Microsoft
- •AWS Cloud Technical Essentials, Amazon
- •Google Cloud Fundamentals, Google