Cantina Senior Staff Software Engineer to own search and recommendation systems end-to-end at Cantina Labs. Responsibilities include architecture, ranking stack design, and operational management of OpenSearch clusters.
Responsibilities
Cantina is a place where people create AI characters and the videos those characters star in, and the catalog grows far faster than anyone can browse. Search and recommendations are what turn that catalog into a product: they decide which characters a new user meets in their first minute, which creators get discovered, and what appears in every feed, leaderboard, and discovery surface we ship. We are hiring a Senior Staff Software Engineer to own that area end to end — retrieval, ranking, relevance quality, and the roadmap for all three. This is the first dedicated hire for search and recommendations, so you will inherit a real production system serving live traffic along with a large amount of unclaimed ground, and you will set the technical direction rather than execute someone else’s.
Own the architecture and roadmap for search and recommendations end to end — query understanding, retrieval, ranking, and the surfaces they feed: search, discovery, the home feed, and our trending and creator leaderboards.
Make relevance a measured discipline rather than an opinion: define the quality metrics per surface, build the offline evaluation and golden-set tooling, and run the online experiments that decide whether a ranking change ships.
Design and evolve the ranking stack — engagement signal pipelines, decay and freshness models, and the pattern we use to combine pre-computed index-time signals with live re-ranking at query time.
Own OpenSearch in production: index and mapping design, reindex and cutover safety, query performance, cost, and the operational headroom of the clusters behind every discovery surface.
Build candidate generation and personalization for recommendation surfaces, and partner with data and ML on the feature pipelines and models behind them.
Work directly with Product and Trust & Safety on what “good” means for each surface — quality gates, cold-start behavior for brand-new creators and characters, and what must never be recommended.
Set the engineering standard for the area: instrumentation, flag-gated rollouts, and runbooks that let other teams change ranking behavior without breaking it.
Act as the technical point of contact for discovery across backend teams, and mentor the engineers whose work touches it.
What you'll bring:
10+ years building production software, with substantial depth in search, ranking, recommendations, or ML-driven relevance on a consumer-scale product.
Hands-on expertise with a production search engine — OpenSearch, Elasticsearch, Lucene, Vespa, or similar — covering index and analyzer design, relevance tuning, and operating the cluster, not only querying it.
Qualification
10+ years building production softwareSubstantial depth in search
Required
10+ years building production software
Substantial depth in search, ranking, recommendations, or ML-driven relevance
