Nikhil Thorat

Career

Databricks' March 2024 announcement says Thorat and Daniel Smilkov each spent a decade at Google before founding Lilac. At Google Brain, Thorat co-created TensorFlow.js and Know Your Data. After Lilac joined Databricks, Thorat worked on AI quality there, and Thorat's GitHub profile now names Anthropic.⁴¹

Open-source work

Thorat's early papers with Smilkov, Fernanda Viégas and Martin Wattenberg focused on tools for understanding models. The Embedding Projector (2016) is an interactive tool for visualizing and interpreting embeddings, and SmoothGrad (2017) reduces noise in gradient-based saliency maps by averaging over noisy copies of an input.⁵⁶

The 2019 TensorFlow.js paper, with Thorat and Smilkov as co-first authors, describes a library for building and running machine learning models in JavaScript in the browser and in Node.js, released under the Apache 2.0 license. Lilac, which Thorat co-founded, is an open-source tool for exploring, clustering and curating text datasets for generative AI, used for tasks such as evaluating model outputs and preparing training data.²³⁴⁷

Work and contributions

• TensorFlow.js Open-source code

  Co-first author, listed with Google Brain, of the 2019 paper describing the design, API, and implementation of TensorFlow.js; Thorat's GitHub profile describes Thorat as a co-creator of the Apache-2.0-licensed library.²¹³

• Lilac Open-source code

  Co-founder, with Daniel Smilkov, of Lilac, an open-source tool for searching, clustering and curating text datasets for language models; the project joined Databricks in March 2024.⁴⁷

Timeline

1. Nov 2016Co-authored "Embedding Projector", an interactive tool for visualizing embeddings.⁵
2. Jun 2017Co-authored "SmoothGrad: removing noise by adding noise".⁶
3. Jan 2019Posted the TensorFlow.js paper as co-first author with Daniel Smilkov.²
4. Mar 19, 2024Lilac, co-founded by Thorat and Daniel Smilkov, joined Databricks.⁴

Other documented roles

Roles the sources document that are not confirmed as current.

• Worked on AI quality, Databricks¹
• Co-founder, Lilac AI¹
• Co-creator of TensorFlow.js and Know Your Data, Google Brain¹²

Where they work

• Anthropic 34 people in the wiki

Related people

• Eugene YanAlso at Anthropic
• Awni HannunAlso at Anthropic
• Rohan TaoriAlso at Anthropic
• Sam McCandlishAlso at Anthropic
• Samuel R. BowmanAlso at Anthropic
• Thierry CruanesDatasets and data
• Bill DolanDatasets and data
• Patrick WendellDatasets and data
• Albert JiangDatasets and data
• Lewis TunstallML frameworks and libraries