Bobak Shahriari
Staff Research Scientist at DeepMind
London, England, United Kingdom
Summary
👤
Senior🎓
Top SchoolBobak Shahriari is a Staff Research Scientist at DeepMind with 11 years of experience blending reinforcement learning, Bayesian decision theory, and practical ML engineering. He leads RL fine-tuning for large and multimodal language models (RLHF, RLAIF) and has contributed to high-profile projects such as Gemini Diffusion and the widely used Acme RL library. His PhD work in Bayesian optimization and statistical modelling produced a popular Python package (Pybo) and informs his approach to automatic parameter tuning and robust decision-making under uncertainty. Bobak’s background spans research and production—he’s implemented RL agents and async update logic in Acme while earlier building data-mining and fraud-detection systems in industry. Based in London, he combines deep theoretical expertise with hands-on engineering to push RL methods from prototypes into scalable tooling.
11 years of coding experience
2 years of employment as a software developer
B. Sc. Honours Physics, B. Sc. Honours Physics at McGill UniversityPhD Computer Science, PhD Computer Science at The University of British ColumbiaM. Sc. Applied Mathematics, M. Sc. Applied Mathematics at Simon Fraser University
English, French
Github Skills (7)
tensorflow10
python10
reinforcement-learning10
documentation9
machine-learning9
algorithms8
jax7
Programming languages (2)
Github contributions (5)
google-deepmind/acme
May 2020 - Oct 2022
A library of reinforcement learning components and agents
Role in this project:
Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:132 commits, 2 PRs, 37 comments in 2 years 4 months
Contributions summary:Bobak primarily contributed to the `acme` library, focusing on reinforcement learning components. Their work involved fleshing out documentation for the MPO loss function and expanding on its configuration options within the `acme/losses/mpo.py`, `acme/agents/dmpo/agent.py`, and `acme/agents/mpo/agent.py` files. They also added and documented examples for running D4PG. The user also modified the asynchronous update logic for variable clients and made some minor modifications to the library structure.
reinforcement-learningreinforcementagentsdeep-reinforcement-learning
bshahr/demos
May 2015 - Oct 2015
Contributions:13 pushes, 1 branch in 5 months
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.