OpenAI Team Update August 2016: New Full‑time Hires and Interns

TL;DR

OpenAI announced on August 16, 2016 that it had hired several new full‑time researchers and engineers and welcomed interns and visitors, strengthening its expertise in speech, safety, strategy, GPU optimization, deep learning, and related fields.

Full‑time hires

OpenAI added five full‑time members: Dario Amodei, Filip Wolski, Jack Clark, Scott Gray, and Zain Shah. Each brings specific background that supports OpenAI’s goals.

Dario Amodei

Dario Amodei was a lead author of Deep Speech 2, a speech system that achieved near‑human performance on many speech tasks, and a main co‑author of "Concrete Problems in AI Safety", which highlights accident risks in machine learning systems. Before OpenAI he worked at Google Brain.

Filip Wolski

Filip Wolski’s recent background is in practical modeling from high‑frequency trading; he previously won the IOI and ACM ICPC programming competitions.

Jack Clark

Jack Clark has spent several years writing about artificial intelligence and distributed systems at Bloomberg and BusinessWeek, covering topics such as memory networks, image generation, reinforcement learning for robots, and diversity in AI. As OpenAI’s Strategy and Communications Director he will handle community outreach, policy, communications, and strategy.

Scott Gray

Scott Gray was an engineer at Nervana Systems where he optimized deep network performance on GPUs; his assembly‑level optimizations for dense linear algebra and convolution remain the fastest available. He also writes software and follows neuroscience research.

Zain Shah

Zain Shah previously led deep learning efforts for a collaborative human‑machine intelligence system at Clara Labs, worked on speech synthesis and computational neuroscience, built Mosaic, founded a mobile behavioral analytics company, and most recently built a GIF search engine using deep multimodal embeddings.

Interns and visitors

OpenAI also welcomed several interns and visitors for a limited period: Catherine Olsson, Harri Edwards, Igor Mordatch, Taco Cohen, and Tambet Matiisen.

Catherine Olsson

Catherine Olsson built OpenAI Gym’s REST API, which already has bindings for Lua, C++, Java, and Rust. She graduated with a perfect GPA in CS and Brain & Cognitive Science from MIT, has research experience in computational neuroscience and psychology, and has taught programming and applied math for six years, including outreach to women and underrepresented minorities.

Harri Edwards

Harri Edwards is a PhD student at the University of Edinburgh researching models that can quickly adapt to new situations by learning to represent datasets.

Igor Mordatch

Igor Mordatch is interested in optimal control, machine learning, and their applications to robotics, biomechanics, and neuroscience. His PhD focused on automated discovery and learning of complex movement behaviors; he will join the faculty at CMU in September 2017.

Taco Cohen

Taco Cohen is a PhD student working on applied and theoretical problems in representation learning; he recently invented group equivariant convolutional neural networks (G-CNNs), a generalization of CNNs that improves statistical efficiency by exploiting symmetries.

Tambet Matiisen

Tambet Matiisen is a PhD student from the University of Tartu, Estonia. He previously worked as a software engineer and founded a startup; his projects include making deep reinforcement learning agents cooperate, predicting a rat’s location from brain activity, and writing an accessible introduction to deep Q‑learning.