Cosmos

AI/ML & recommendation systems

Lead AI/ML and recommendation systems at Cosmos (4M+ users), building core discovery products from 0→1, including the For You feed, text and color search, similar elements, and collection recommendations. Designed and implemented additional AI features including autocomplete and research-driven AI captioning, while contributing to backend systems including element processing, reliability, and systems design.

As the platform scaled, led the next phase of the recommendation stack, designing the data warehouse and real-time feature layer, building ingestion pipelines across engagement signals such as clicks, views, likes, and slates, and training in-house models to power recommendation and ranking systems across the product.

Features

Cosmos × Qdrant article featuring an interview I had with Qdrant Vector Database on our recommendation systems.

AI/ML work has also been covered by Fast Company and Business of Home, with broader Cosmos features in WWD and CULTURED.

Recognition

2026 Webby nominee for Best Use of AI & Machine Learning and People’s Voice Winner for Best Visual Design

Fast Company Innovation by Design finalist

Visit Cosmos

A few things I’ve been working on:

AI captions. Identifying the creator, subject, style, and context of images to support attribution and discovery. 77.5K views on X
Collection recommendations. Finding related elements to build out a collection. 22.9K views on X
Autocomplete. Suggestions drawn from real searches, with a cached trie-based system responding in under 10ms. 5.4K views on X
Similar elements. More related images, right below the one you’re viewing. 5.1K views on X
Color extraction. Comparing RGB with perceptual LAB color detection and ΔE2000 merging to capture more accurate, distinct colors. 2.7K views on X

View counts as of Sept. 2026.