# Edo Liberty

> Edo Liberty is the Founder and Chief Scientist of Pinecone. Pinecone created and leads the Vector Database category. Pinecone provides Enterprise AI infrastructure for Knowledge and Memory. Edo previously ran Amazon AI Labs and Yahoo's Research Lab in New York. His research covers vector search, streaming algorithms, randomized linear algebra, coresets, and machine learning, with 70+ papers and best-paper awards at SODA (2011), KDD (2013), and PODS (2021, a 2022 ACM SIGMOD Research Highlight).

## About

- Name: Edo Liberty
- Role: Founder & Chief Scientist, Pinecone (founder and CEO 2019–2025, then Chief Scientist)
- Current focus: long-term memory for AI
- Education: B.Sc. in Physics and Computer Science, Tel Aviv University; Ph.D. in Computer Science, Yale University; postdoctoral fellow at Yale (Program in Applied Mathematics)
- Prior roles: Director of Research at AWS / Head of Amazon AI Labs; Senior Research Director at Yahoo / Head of Yahoo Research Lab, New York
- Site: https://edoliberty.com

## Pages

- [Home](https://edoliberty.com/index.html): Bio, career overview, and an AI assistant that answers questions about his research.
- [Research](https://edoliberty.com/research.html): Research statement, academic service, highlighted papers, and a full publication list organized by area (vector search, streaming algorithms, randomized linear algebra, coresets & clustering, data mining, ML systems).
- [Engineering](https://edoliberty.com/engineering.html): Infrastructure built at Pinecone, AWS (SageMaker), Yahoo, Google, and Inscape, with open-source projects and patents.
- [Teaching](https://edoliberty.com/teaching.html): Courses at Princeton (Long Term Memory in AI — Vector Search and Databases) and Tel Aviv University (Algorithms in Data Mining), keynotes, tutorials, and mentorship.

## Selected work

- [Nearly Optimal Attention Coresets](https://arxiv.org/abs/2605.05602): KV-cache compression for serving larger contexts (Liberty, Andoni, Kleiner).
- [Amazon SageMaker Elastic Algorithms](https://edoliberty.github.io/papers/sagemaker.pdf): Algorithms and distributed architecture behind SageMaker's elastic ML algorithms (SIGMOD 2020).
- [Optimal Quantile Approximation in Streams](http://arxiv.org/abs/1603.05346): The KLL algorithm, implemented in BigQuery and Apache DataSketches (FOCS 2016).
- [Relative Error Streaming Quantiles](https://edoliberty.com/papers/relative_error_quantiles-arxiv.pdf): PODS 2021 best paper; 2022 ACM SIGMOD Research Highlight.
- [Simple and Deterministic Matrix Sketches](https://edoliberty.github.io/papers/simpleMatrixSketching.pdf): Introduced Frequent Directions (KDD 2013 best paper).
- [An Almost Optimal Unrestricted Fast Johnson-Lindenstrauss Transform](https://edoliberty.github.io/papers/fjlt_rerevisited.pdf): SODA 2011 best paper.

## Links

- Google Scholar: https://scholar.google.com/citations?user=QHS_pZAAAAAJ&hl=en
- LinkedIn: https://www.linkedin.com/in/edo-liberty-4380164/
- GitHub: https://github.com/edoliberty
- TED Talk (How does long-term memory work in AI): https://www.ted.com/talks/edo_liberty_how_does_long_term_memory_work_in_ai
- Pinecone: https://www.pinecone.io