01 / VSCO · 2024–Present
ML-Driven Creative Tools & Feed Recommendation
7%
Engagement & retention lift
5%
Trial conversion
0→1
First ML feed rec system at VSCO
The insight
Photographers long for community. They shoot, edit, and post, but mostly in isolation. VSCO had built some of the best creative tools in the world, yet the experience stopped at the edge of the individual. There was no loop connecting what users discovered to what they created. Inspiration happened elsewhere, on Instagram, Pinterest, in DMs, and then users returned to VSCO to edit. That gap was the opportunity.
Closing the loop
I led two connected workstreams as part of a single platform thesis: make discovery smarter, then make the jump from inspiration to creation effortless. The first was VSCO's 0→1 ML feed recommendation system on iOS, the first time the app could surface content that actually matched a user's aesthetic, not just chronological posts. I partnered directly with ML engineers to define training data, offline evaluation metrics, and online A/B test design, and made targeted contributions to the iOS codebase to prototype changes before committing engineering resources.
The second was a suite of AI-powered editing tools built on external models, where I owned model selection, integration constraints, evaluation criteria, and rollout strategy end-to-end. The goal was the same: when a photographer sees something that moves them, VSCO should be the place where they can immediately act on it. Discover. Get inspired. Create. That's the loop.
Building the foundation
Alongside both workstreams, I built and now teach VSCO's experimentation and analytics framework, shifting the culture from intuition-driven to evidence-driven decision making across PM and engineering teams. I also led vendor negotiations, data platform modernization, and co-authored VSCO's AI ethics standards and model evaluation best practices with leadership.
