Projects
I build to learn. Each of these started as a real problem I wanted to solve — and became a lesson in making product decisions under constraint.
Spexel
AI · MedTechUpload an engineering drawing. Get a traceability matrix in seconds.
Try it at spexel.io →
Problem
Engineers in manufacturing and regulated industries spend hours manually reading engineering drawings to extract specifications into traceability matrices. It's tedious, error-prone, and a bottleneck on every product release cycle — especially when drawing packages span hundreds of pages across multiple revisions.
Key Decisions
The core product decision was accuracy over speed: a single-model approach was inconsistent on domain-specific notation like GD&T callouts and fit tolerances. I designed a two-stage pipeline — custom OCR for deterministic text extraction, followed by LLM validation that interprets specs in visual context. The tradeoff was latency; the gain was reliability. I also chose a privacy-first architecture with zero server-side storage, which became a key differentiator for regulated-industry customers.
What Shipped
A production web app at spexel.io — upload a drawing (PDF, PNG, JPEG), review AI-extracted specifications, customize the matrix, and export. Multi-page drawing support, per-page editing, and a clean review UI so engineers can catch and correct errors before export. GDPR-compliant with zero data retention by default.
What I Learned
Talking to engineers early would have saved several pivots. The most valuable feedback loop was watching someone use it on a real drawing for the first time — the gaps between what I built and what they needed were immediately obvious. Ship early, watch real usage, iterate.
UltimateMe
iOS · HealthAI health coaching app that connects wearable data to personalized guidance.
Problem
Existing health apps surface data but don't do anything useful with it. Your watch knows you slept poorly and skipped the gym — but it never connects those dots or tells you what to do next. There's a gap between data collection and actionable coaching.
Key Decisions
I chose Supabase over local-only storage early on, even though local would have been faster to ship. Cross-device sync and any future social features would require a cloud layer — building it later would have meant re-architecting around user data, which is painful and risky in a health context. I also chose HealthKit as the single source of truth rather than asking users to manually log anything, which removed the biggest friction point in health apps.
What Shipped
A native iOS app that reads from HealthKit — sleep, activity, heart rate, HRV — and pipes that context into Gemini AI to generate personalized coaching responses. Cloud sync via Supabase, GDPR-compliant data handling, and 57 unit tests covering the core coaching logic.
What I Learned
Shipping a regulated-adjacent health product means data privacy isn't a feature you add later — it's an architectural constraint you design around from day one. Every schema decision, every permission prompt, every sync event has compliance implications. I learned to think about privacy the way I think about quality in MedTech: it has to be built in, not bolted on.
Congressional Trade Tracker
Automation · CivicAutomated monitor that surfaces congressional financial disclosures in real time.
Problem
Congressional trading disclosures are legally required to be public — but they're buried in PDFs on government websites, filed inconsistently, and effectively inaccessible to anyone who isn't actively watching. The transparency exists on paper, but not in practice.
Key Decisions
I chose Discord over email for delivery — not for technical reasons, but for distribution ones. The community of people who care about this data already lives in Discord servers. Meeting users where they are, rather than asking them to check a dashboard or subscribe to a newsletter, turned out to be the most important product decision I made.
What Shipped
A Python bot that scrapes disclosure PDFs on a schedule, parses the trade data, stores it in SQLite, and fires formatted alerts to Discord in real time whenever new trades are detected. Tracks multiple politicians simultaneously and handles deduplication across runs.
What I Learned
Distribution strategy matters as much as the product itself. I spent more time on the parsing logic than the delivery mechanism — but the Discord integration drove engagement more than any feature I added. The lesson: figure out where your users already are and bring the product to them, not the other way around.