Feng Niu

Feng Niu is a computer scientist and company founder whose work spans parallel optimization, statistical inference at scale and production machine-learning systems. He completed a computer sciences PhD at the University of Wisconsin–Madison in 2012, where he was first author of the 2011 Tuffy paper and co-wrote Hogwild!, a lock-free way to run stochastic gradient descent in parallel that received the NeurIPS Test of Time Award in 2020. Alation (2012) and Lattice Data (2015) followed as companies he co-founded; Apple acquired Lattice in 2017, and he worked there as a research scientist until 2020. In 2019, he coauthored Overton, a data system for building and monitoring production machine-learning applications. Since 2020, he is CEO and co-founder of Evidently, an AI platform for hospitals and health systems.

Now at: Evidently
Badges: Founder
Bridge

Timeline

2008–2012

Wisconsin PhD under Ré and Doan

He studies computer sciences at the University of Wisconsin–Madison from 2008, co-advised by Christopher Ré and AnHai Doan, and finishes his PhD in 2012. In 2011, he is first author of Tuffy, written with Ré, Doan and Jude Shavlik, which scales up statistical inference in Markov logic networks using a relational database system.
2011

Hogwild! at NeurIPS

With Benjamin Recht, Christopher Ré and Stephen Wright, members of the optimization group at the Wisconsin Institute for Discovery, he writes Hogwild!, published at NeurIPS 2011. Each worker runs stochastic gradient descent on a different subset of the data and updates the shared model parameters asynchronously without locking, which gives a near linear speedup on sparse problems.
2012

Co-founding Alation

He is one of the four co-founders of Alation, the Silicon Valley data-catalog company founded in 2012 with Satyen Sangani, Aaron Kalb and Venky Ganti.
2015

Co-founding Lattice Data

In 2015, Lattice Data is founded to commercialize DeepDive, a system created at Stanford to extract value from dark data, and he is one of its four co-founders with Christopher Ré, Michael Cafarella and Raphael Hoffmann. He serves as the company's chief engineering officer.
2017–2020

Apple and the Overton system

Apple acquires Lattice Data in 2017 for a reported $200 million, and its 20 engineers join the company. He works at Apple as a research scientist from 2017 to 2020. In 2019, he coauthors Overton with Christopher Ré, Pallavi Gudipati and Charles Srisuwananukorn, a data system for building, monitoring and improving production machine-learning applications that has answered billions of queries.
2020

Becoming CEO of Evidently

In 2020, he becomes CEO and co-founder of Evidently, a company that builds AI tools for hospitals and health systems.
2020

A NeurIPS Test of Time Award

NeurIPS gives its 2020 Test of Time Award to Hogwild!, which he wrote with Benjamin Recht, Christopher Ré and Stephen Wright. The award recognizes the first paper to show how to parallelize stochastic gradient descent without any locking mechanism while achieving strong performance guarantees.
As of October 2026

Evidently's CEO and a Series A

He is CEO and co-founder of Evidently, an AI platform that synthesizes fragmented medical data from disparate sources into actionable insights for hospitals and health systems. In January 2025, DN Capital announces an oversubscribed $15 million Series A investment in Evidently, backing the team he runs as CEO alongside Chief Product Officer Kalie Dove-Maguire.