Ahmad Beirami
Ex-Research Leader @ Google (DeepMind), Meta, & EA
PhD / Postdoc @ MIT, Harvard, Duke, & Georgia Tech
Ahmad Beirami is a technologist passionate about building innovative, efficient, and scalable products. With over 15 years of experience in AI, more than 5 years of hands-on leadership developing large language model (LLM) products at Google and Meta, and building an award-winning AI agent framework at Electronic Arts, he bridges deep technical expertise with real-world product impact. His recent work on LLM alignment advanced the Pareto frontier of Gemini model capabilities and earned an Outstanding Paper Award at ICLR 2025. See Academic Engagements for his academic contributions.
Selected Product Achievements
Ahmad Beirami has led dozens of product launches that span RL in games, integrity of video games, LLM safety/robustness/fairness, RL in LLMs, and LLM efficiency.
- Google DeepMind: Pushed the Gemini capabilities–compute Pareto frontier by leading the cross-org launches of controlled decoding and InfAlign for RL post-training and test-time compute scaling.
- Google Research: Delivered 2× inference speedups for Gemini and Gemma via speculative decoding.
- Meta AI: Led a cross-org effort to invent and prototype CAIRaoke, Meta’s first tool-using digital assistant.
- Electronic Arts: Delivered EA’s first RL-driven playtesting, accelerating pre-release bug discovery and balance tuning.
Selected Awards & Honors
- Outstanding Paper Award, International Conference on Learning Representations (2025).
- Outstanding Paper Award, IEEE Computational Intelligence Society (2022).
- Sigma Xi Best PhD Thesis Award, Georgia Tech (2015).
- Bronze Medal, Iranian Mathematics Olympiad (2002).