I spent three years validating pharmaceutical systems under FDA rules. Now I build products, retrieval systems, and models with the same discipline.

Before pivoting into product engineering, I worked as a validation engineer at cGMP Consulting Inc. in Illinois, with client engagements at AbbVie, Fresenius Kabi, and other pharmaceutical and medical device firms.

The work was GMP-heavy: IQ/OQ/PQ protocols, Periodic Validation Reviews, FMEAs, SOPs, change controls, and CAPA support under FDA 21 CFR. It trained me to document precisely, prioritize risk, and treat quality gates as part of shipping — habits I still use when designing RLS policies, evaluation harnesses, and pre-launch security checks.

I started in full-stack web development — React frontends, Supabase backends, billing flows, and the documentation that makes products maintainable. Over time that grew into AI integration: RAG pipelines, agent tooling, and cost-conscious LLM features in Edge Functions.

Today I work across the stack — from schema design and PWAs to GraphRAG backends and structured authz testing. I'm also pursuing an MS in Data Science at the University of Pennsylvania, which pushes me to pair shipping instinct with rigorous evaluation and modeling fundamentals.

My early projects were small tools and experiments — automating repetitive workflows, learning how to structure databases, and figuring out why security matters before you have users (and especially after).

That curiosity carried into production work: retail operating systems for Latin America, parking platforms with an AI concierge layer, and pre-launch authz review on multi-tenant Supabase apps.

Outside of client and product work, I build small tools that scratch a real itch — usually at the intersection of automation, data, and creative tooling. They're not side projects for a portfolio checkbox; each one started from a question I actually wanted answered.

The Prediction Market Forensic Monitor came from watching prediction markets around major news events: wallets placing large bets minutes before headlines broke, patterns that look a lot like privileged information. I wanted an n8n pipeline that could flag those irregularities early — scrape activity, score timing against public events, and surface leads worth a closer look.

Brand Sentinel started closer to home. My mom has a startup idea tied to my grandparents' farm, and she needed a way to track what competitors were doing without spending hours on manual research. The workflow scrapes competitor sites, runs summaries through Gemini and OpenAI, and lands structured updates in Airtable — a lightweight market watch for a family business finding its footing.

The ComfyUI sprite pipeline is the creative bet: three workflows (poses, walking, running) that turn character LoRAs into multi-directional chibi frames with ControlNet and pixel-art styling. The goal is to see whether I can solo-develop an entire indie game with AI handling the asset grind — so I can spend my time on mechanics, story, and feel instead of redrawing every frame by hand.