Gal Dalal  

 



  ![](/sites/default/files/person/Gal.Dalal_.jpg)

  

 Gal Dalal is Senior Research Scientist working on Reinforcement Learning (RL) theory and applications at NVIDIA Research. Previously, he co-founded Amooka-AI, which later became Ford Motor Company’s L3 driving policy team. He obtained his BSc in EE from Technion, Israel, summa cum laude, and his PhD from Technion as a recipient of the IBM fellowship. Gal interned at Google DeepMind and IBM Research, and received the 2019 AAAI Best (“outstanding”) Paper Award, ranked 1st among 1150 accepted papers.

Gal’s research interests span over both RL theory and applications. His theory work includes time-delayed decision making, multi-step greedy policies, and finite-time convergence of RL algorithms. On the application side, he worked on real-world problems such as autonomous driving, GPU cache control, network congestion control, datacenter cooling, and smart-grid management.



   Research Area(s)

[Artificial Intelligence and Machine Learning ](/index.php/research-area/machine-learning-artificial-intelligence)

[Autonomous Vehicles](/index.php/research-area/autonomous-vehicles)

 

 

  

 Main Field of Interest

[Artificial Intelligence and Machine Learning ](/index.php/research-area/machine-learning-artificial-intelligence)

 

  

 Google Scholar

[https://scholar.google.com/citations?user=NfJiSMMAAAAJ&hl=en&oi=ao](https://scholar.google.com/citations?user=NfJiSMMAAAAJ&hl=en&oi=ao)

 

  

 

 

 



 ### Publications

 

### 2025 

[SoftTreeMax: Policy Gradient via tree expansion ](/index.php/publication/2025-02_softtreemax-policy-gradient-tree-expansion)

[Gal Dalal](/index.php/person/gal-dalal), [Assaf Hallak](/index.php/person/assaf-hallak), Gugan Thoppe, [Shie Mannor](/index.php/person/shie-mannor), [Gal Chechik](/index.php/person/gal-chechik)



[ICML 2025](https://icml.cc/virtual/2025/poster/43515)









### 2023 

[Implementing Reinforcement Learning Datacenter Congestion Control in NVIDIA NICs](/publication/2023-01_implementing-reinforcement-learning-datacenter-congestion-control-nvidia-nics)

Benjamin Fuhrer, Yuval Shpigelman, [Chen Tessler](/person/chen-tessler), [Shie Mannor](/person/shie-mannor), [Gal Chechik](/person/gal-chechik), Eitan Zahavy, [Gal Dalal](/person/gal-dalal)



[CCGrid 2023](https://arxiv.org/abs/2207.02295)









### 2022 

[On Covariate Shift of Latent Confounders in Imitation and Reinforcement Learning](/publication/2022-10_covariate-shift-latent-confounders-imitation-and-reinforcement-learning)

Guy Tenneholtz, [Assaf Hallak](/person/assaf-hallak), [Gal Dalal](/person/gal-dalal), [Shie Mannor](/person/shie-mannor), [Gal Chechik](/person/gal-chechik)



[ICLR](https://iclr.cc/)









[Reinforcement Learning for Datacenter Congestion Control](/index.php/publication/2022-02_reinforcement-learning-datacenter-congestion-control)

[Chen Tessler](/index.php/person/chen-tessler), Yuval Shpigelman, [Gal Dalal](/index.php/person/gal-dalal), Amit Mendelbaum, Doron Kazakov, Benjamin Fuhrer, [Gal Chechik](/index.php/person/gal-chechik), [Shie Mannor](/index.php/person/shie-mannor)



IAAI 2022









### 2021 

[Improve Agents without Retraining: Parallel Tree Search with Off-Policy Correction](/publication/2021-10_improve-agents-without-retraining-parallel-tree-search-policy-correction)

[Gal Dalal](/person/gal-dalal), [Assaf Hallak](/person/assaf-hallak), [Steven Dalton](/person/steven-dalton), [Iuri Frosio](/person/iuri-frosio), [Shie Mannor](/person/shie-mannor), [Gal Chechik](/person/gal-chechik)



[Advances in Neural Information Processing Systems 34 (NeurIPS 2021)](https://nips.cc/Conferences/2021/)









[Planning and Learning with Adaptive Lookahead](/index.php/publication/2021-01_planning-and-learning-adaptive-lookahead)

Aviv Rosenberg, [Assaf Hallak](/index.php/person/assaf-hallak), [Shie Mannor](/index.php/person/shie-mannor), [Gal Chechik](/index.php/person/gal-chechik), [Gal Dalal](/index.php/person/gal-dalal)



[Arxiv](https://arxiv.org/abs/2201.12403)









[Acting in Delayed Environments with Non-Stationary Markov Policies](/index.php/publication/2021-01_acting-delayed-environments-non-stationary-markov-policies)

Esther Derman, [Gal Dalal](/index.php/person/gal-dalal), [Shie Mannor](/index.php/person/shie-mannor)



ICLR 2021









### 2020 

[The Architectural Implications of Distributed Reinforcement Learning on CPU-GPU Systems](/index.php/publication/2020-12_architectural-implications-distributed-reinforcement-learning-cpu-gpu-systems)

Ahmet Inci, Evgeny Bolotin, [Yaosheng Fu](/index.php/person/yaosheng-fu), [Gal Dalal](/index.php/person/gal-dalal), [Shie Mannor](/index.php/person/shie-mannor), [David Nellans](/index.php/person/david-nellans), Diana Marculescu



[Workshop on Energy Efficient Machine Learning and Cognitive Computing (EMC2)](https://www.emc2-ai.org/virtual-20)