Engineering leader & builder.
I like turning messy, ambiguous problems into real products and teams that can keep shipping after I leave the room. Most recently I was a founder & CTO building AI for sustainability — measuring the carbon footprint of compute — where I took the product from 0 → 1 and built a 6-person international engineering team from scratch, with an emphasis on scalability for both the system and the people in it.
Before that I spent a decade as an engineering leader and ML engineer at Aurora (autonomous driving), Uber, Outreach (founding ML engineer), and Sony in Tokyo — building distributed training pipelines, data platforms, and the ML infrastructure underneath them. Outside of work I build small web apps when an idea won't leave me alone.
I'm always up for a conversation — about engineering leadership, sustainability, AI products, side projects, or anything else you think I'd find interesting. Reach me at:
Built a platform that measures the carbon footprint of compute — from chips to data centers — including a domain-specific fine-tuned LLM and an email agent for automated data collection. Sold into hyperscalers; published the model architecture at an international conference.
Led the data & annotation ML stack: a scalable annotation workflow with quality scoring, an object-detection-based validation framework that catches label defects pre-delivery, Kubeflow pipelines for end-to-end ML lifecycle, and an unsupervised visual search system over millions of driving logs.
Built a distributed multi-GPU training pipeline for DeepLab v3 with hyperparameter tuning and live monitoring, and a scalable inference system on TensorFlow Serving. Published and benchmarked computer vision models on AWS Marketplace.
Built the company's first ETL pipelines and data warehouse, an intent classification model for prospect responses, and ML-driven APIs surfacing content-effectiveness insights into the core product. Automated A/B testing on Spark + Airflow.
NLP and named-entity recognition features, recommendation systems to reduce churn, and statistical analysis on user behavior powering targeted advertising and segmentation.
Photos via Unsplash.
Web app · End-to-end encrypted notes
Notes encrypted in your browser before they ever reach the server. Username and password only — no email, no identity check.
Why: I wanted a place I could actually trust with my most private notes.
Web app · Family activity planner
A planner that answers “where should I take my kid today?” — kid-friendly activities and destinations near you.
Why: as a new mom, I struggled to find places that worked for both of us.
Chrome extension · Intentional browsing
A pause before distracting sites — today's timeline and a 7-day chart let you decide instead of drifting. All local, no analytics.
Why: my time is limited and I get distracted easily.