captain
The open research lab for retail product discovery
- Pre-training small session transformers for retail revenue lift
- Gradient-boosted trees vs. transformers for product ranking
- Correcting for Pollution: How training rankers on clicks overfits old models
- An AI research agent to maintain on-site search results
- A/B testing product search engines for retail
Search that self-improves.
Fast, accurate product and file search. Offload tuning without losing control. Always in sync with your sources.
Understanding how shopper intent improves product discovery.
We study which search and ranking methods lead to more sales, and publish what we find for the retail AI and ML community.
Open research
Studies on public retail data and novel methodologies.
Practical guides
Get more from the engine you run, or build your own.
Retail partners
Partnering with industry AI and ML retail teams to understand the revenue implications of better retrieval and ranking.
Founded by Lewis Lien Polansky (CEO) and Edgar Babajanyan (CTO).
Edgar did NLP and AI research at Purdue and now builds the lab's search systems. Lewis came to search from cybersecurity and runs retail partnerships and developer relations. They also built and support filesearch.dev, a file search RAG engine with frontier accuracy.
Captain is backed by Y Combinator (W26).