Work

At Liftoff, I lead AI creative generation across video, banner, and HTML formats—pipelines that combine LLMs, vision models, image and video generation, and TTS so briefs turn into shippable ads with human approval as the main bottleneck. The same role spans outbound sales agents, an agentic dashboard for ad ops, and the observability and eval layers that keep all of it honest in production.

At WOMBO, I worked backend and ML production for a consumer generative-AI app at real scale: a queue-based serving stack that cut new-model time-to-ship roughly fivefold, open-source diffusion and video models in production, GPT-4 in the text-to-image flow for prompts, and a CLIP + Pinecone search service over millions of user images— plus the usual cost and infra tightening (caching, autoscaling GPU clusters, storage reclamation jobs).

At Google, I spent five years as an SWE (L3→L4): YouTube Music recommendation and experiments in C++, an internationalized Skia image pipeline for shareable link previews, livestream notification quality, and earlier Assistant features and Android prototypes in Google's internal C++ framework—plus the engineering residency rotations that got me started on production systems at absurd scale.