Skip to content

Search and ranking

Search and recommendations that surface the right thing.

When search, feeds or notifications show the wrong results, people stop using them. We audit relevance with judged queries, improve retrieval and ranking and confirm the gain in live experiments.

Who it's for

Product teams whose search, recommendations, feeds or notifications don't surface the right thing.

What we deliver

  • A relevance audit with judged queries.
  • Hybrid retrieval.
  • Rerankers and learning-to-rank.
  • Personalization.
  • Offline metrics tied to online A/B tests, so improvements are real.

How we measure it

  • nDCG, MRR and recall offline.
  • Click-through, conversion and retention in experiments.

Background

And led the recommendations and personalization machine learning team at Quizlet, was Head of Recommendations Platform and Science at Etsy and worked on large-scale experimentation at Pinterest. The RAG retrieval quality benchmark below reports the same ranking metrics, such as MRR, used to judge search.

Related work

Built in the open

Questions

Common questions

How do you know the AI is working?
Each project gets a test set from your real work and agreed metrics (for example, answer correctness for an assistant or hours saved for an automation), measured before and after every change.
Which AI models do you use?
Claude, OpenAI models and open-source models, chosen per task. The test set decides, including cost.
Who will I work with?
And, directly, from the first call to handoff. There's no handoff to a junior team.

Have a process that eats hours every week? Let's look at it together.