backend engineer - billing, monetization, platform integrity

I'm a backend engineer at Luma AI, where I work on billing, monetization, and platform integrity. Previously I built payments and crypto infrastructure, including an open-source Stripe library for React Native that has processed $1M+ in transactions for more than 32k customers.
Luma AI
Backend Engineer, Feb 2026 - present
Billing, monetization, and platform integrity for the agent platform. I built the spending cap at the reserve gate, which has stopped 4,178 teams at the cap they chose - median usage at refusal 1.01x the cap, against 2.5x-4.9x before - and fixed settlement correctness at scale, restoring 8.8M credits of paid overage to 2,191 teams. I cut long database connection holds in agent jobs from 567k a day to 18, built the API content-policy gate that has blocked 523k prompts before generation (12% of all attempts) and the phone-verification signup gate that has stopped 8.8k abusive signups, and shipped the UI directives framework now carrying 3.08M calls a week from 185k users - 18 other engineers have built on it.
Built with scaffold-eth / BuidlGuidl: a meta multi-signature wallet (repo), a streaming-enabled multi-signature wallet for recurring payments (repo), and Push Button, a template for multiplayer turn-based games with commit-reveal randomness.
dont-export-your-pk, use-redux-effect, react-native-form-inputs, and pageview_parallax.
Multilingual Toxic Comment Classification
Kaggle competition: train on English toxic comments, detect toxicity across languages. With little usable multilingual data, we generated our own with translation models, trained language-specific models on the translated sets, and blended the results over many rounds of hyperparameter tuning. Team Sentimental took a silver medal, 14th of 1600+ teams, and my Kaggle profile was promoted to Competitions Expert.