About
Main bio
I spend my days building products where “just figure it out” is not good enough. Financial risk, health records, training plans, and football forecasts all become difficult when useful signals are buried inside noise. My job is to make the important part obvious.
I've worked across SwiftUI, FastAPI, Next.js, Supabase, PostgreSQL, and self-hosted infrastructure. Sometimes there is a complete dataset and sometimes there is only a rough idea. What stays consistent is asking good questions, learning the system, and shipping something that can survive real use.
Currently, my main focus is Clavix, a personalized financial-news and portfolio-risk product. I also build Concord as a private health companion, maintain VarFoot, and keep testing a World Cup forecasting model against live markets.
it's me
MY BUILDING ORIGIN
I started with Walter
I started coding in fifth grade because I wanted a robot to walk, listen, speak, and have a mind of its own. Neither Walter exists anymore, so the archive is the evidence.
fifth grade, first steps.
Raspberry Pi 3 / 8 servos
The first Walter walked
A Raspberry Pi 3 and eight servos powered two legs. Walter walked, listened, spoke, performed motions, searched the web, and played music before ChatGPT existed.
Walter V3 learned a personality
I build across the whole stack: SwiftUI on the front, FastAPI and Supabase on the back, self-hosted in production. The interesting part is connecting the pieces so the whole thing actually runs, not just the screen you can see.
Orange Pi 5 Plus / local 2B model
make the important part obvious.
What the robots taught me
Make complicated systems understandable, keep the important feedback visible, and build every layer until the whole thing works outside the demo.
MY BUILDING ORIGIN
Timeline
Clavix
Building the iOS app, FastAPI backend, scoring pipeline, subscriptions, and self-hosted production system for personalized market news and portfolio risk.
2025 - current
Concord
Exploring how HealthKit and Health Connect data, careful symptom capture, and an AI companion can become a useful doctor-ready summary without pretending to diagnose.
Active side project
WC2026 Predictor
Combining Dixon-Coles score modeling with XGBoost, public football data, calibration checks, and live market comparisons against the betting market.
Research build
VarFoot
Built an AI-supported soccer roadmap that assesses a player, compares the result with varsity expectations, and turns the gap into a weekly training plan.
LexHack 2026