CV

Education

• M.S. in Computer Science, University of Washington (Paul G. Allen School of Computer Science & Engineering), 2025 – present
• B.S. in Computer Engineering with Distinction, University of Washington (Paul G. Allen School of Computer Science & Engineering), 2025
• B.A. in Comparative History of Ideas, University of Washington (College of Arts and Sciences), 2025

Research experience

• Jun. 2026 – present: Graduate Student Researcher
  • Allen Institute for AI (Ai2), AllenNLP Scaling Team
  • Integrating multimodal capabilities into OLMo-core for mixture-of-experts LLMs
  • Raised MMMU from 64.92% to 81.75% and CharXiv from 33.00% to 40.60% on 4B and 12B MoE OLMo models
  • Improved OLMo-eval infrastructure to support multimodal evaluation benchmarks
  • Increased training throughput 8x in wall-clock time and 4x in GPU efficiency
• Sep. 2025 – present: Graduate Student Researcher
  • Ai2 PRIOR/Robotics & RAIVN Lab, University of Washington
  • Advisor: Prof. Ranjay Krishna
  • Core contributor to MolmoAct2; built the data-processing and tokenizer pipeline
  • Contributed to MolmoSpaces; built the Isaac Sim conversion pipeline
  • Training a G1 humanoid robot using motion retargeting derived from human video data
• Mar. 2025 – present: Graduate/Undergraduate Researcher
  • Social RL Lab, University of Washington
  • Advisor: Prof. Natasha Jaques
  • Working on self-play RL for generating secure code
  • Steered state-of-the-art LLMs toward human-like behavior using multi-turn RL (OpenRLHF)
• Nov. 2024 – Mar. 2025: Undergraduate Researcher
  • ZLab & RSE Lab, University of Washington
  • Advisors: Prof. Luke Zettlemoyer & Prof. Dieter Fox
  • Investigated internet-scale, unsupervised learning for complex robotic tasks in diverse environments
  • Fine-tuned video-diffusion models to generate robot trajectory videos in a RoboCasa-like environment
• Jan. 2024 – Aug. 2025: Honors Undergraduate Researcher
  • ACME Lab (HLS4ML Group), University of Washington
  • Advisor: Prof. Scott Hauck
  • Authored an honors thesis benchmarking automated neural network hardware acceleration
  • Implemented a neural network in SystemVerilog to compare against HLS code

Industry experience

• Jun. 2024 – Sep. 2024: Software Engineer Intern
  • Amazon, AWS SageMaker (Seattle)
  • Architected CI/CD infrastructure using the AWS Cloud Development Kit to automate testing and releases
  • Automated release workflows via conda-forge, eliminating 2–3 weeks of manual work per release
  • Automated AWS-based tests for every contribution to the open-source repository

Skills

• Languages: Python, Java, C/C++, Verilog/SystemVerilog, Rust, Bash
• Frameworks & ML systems: PyTorch, CUDA, vLLM, DeepSpeed, OpenRLHF, OLMo-core, open-instruct
• Robotics & simulation: NVIDIA Isaac Sim/Lab, MuJoCo, RoboCasa
• Infrastructure & developer tools: AWS (CDK, SageMaker), Slurm, Docker, Linux, Git, Conda

Publications

Demystifying Reinforcement Learning Post-Training of Language Models

Clay, D., et al.Donovan Clay, Saket Gollapudi, Sankar Harilal, Min Jang, Jacob Morrison, Sewoong Oh, Natasha Jaques (2026). "Demystifying Reinforcement Learning Post-Training of Language Models." Under review at the International Conference on Learning Representations (ICLR 2027).

MolmoAct2: Action Reasoning Models for Real-world Deployment

Fang, H., Duan, J., Clay, D., et al.Haoquan Fang, Jiafei Duan, Donovan Clay, Sam Wang, Shuo Liu, Weikai Huang, Xiang Fan, Wei-Chuan Tsai, Shirui Chen, Yi Ru Wang, Shanli Xing, Jaemin Cho, Jae Sung Park, Ainaz Eftekhar, Peter Sushko, Karen Farley, Angad Wadhwa, Cole Harrison, Winson Han, Ying-Chun Lee, Eli VanderBilt, Rose Hendrix, Suveen Ellawela, Lucas Ngoo, Joyce Chai, Zhongzheng Ren, Ali Farhadi, Dieter Fox, Ranjay Krishna (2026). "MolmoAct2: Action Reasoning Models for Real-world Deployment." Conference on Robot Learning (CoRL 2026). Spotlight (top 10%).

WildDet3D: Scaling Promptable 3D Detection in the Wild

Huang, W., ..., Clay, D., et al.Weikai Huang, Jieyu Zhang, Sijun Li, Taoyang Jia, Jiafei Duan, Yunqian Cheng, Jaemin Cho, Matthew Wallingford, Rustin Soraki, Chris Dongjoo Kim, Shuo Liu, Donovan Clay, Taira Anderson, Winson Han, Ali Farhadi, Bharath Hariharan, Zhongzheng Ren, Ranjay Krishna (2026). "WildDet3D: Scaling Promptable 3D Detection in the Wild." Robotics: Science and Systems (RSS 2026).

MolmoSpaces: A Large-Scale Open Ecosystem for Robot Navigation and Manipulation

Kim, Y., ..., Clay, D., et al.Yejin Kim, Wilbert Pumacay, Omar Rayyan, Max Argus, Winson Han, Eli VanderBilt, Jordi Salvador, Abhay Deshpande, Rose Hendrix, Snehal Jauhri, Shuo Liu, Nur Muhammad Mahi Shafiullah, Maya Guru, Arjun Guru, Ainaz Eftekhar, Karen Farley, Donovan Clay, Jiafei Duan, Piper Wolters, Alvaro Herrasti, Ying-Chun Lee, Georgia Chalvatzaki, Yuchen Cui, Ali Farhadi, Dieter Fox, Ranjay Krishna (2026). "MolmoSpaces: A Large-Scale Open Ecosystem for Robot Navigation and Manipulation." Robotics: Science and Systems (RSS 2026).

Consistently Simulating Human Personas with Multi-Turn Reinforcement Learning

Abdulhai, M., Cheng, R., Clay, D., et al.Marwa Abdulhai, Ryan Cheng, Donovan Clay, Tim Althoff, Sergey Levine, Natasha Jaques (2025). "Consistently Simulating Human Personas with Multi-Turn Reinforcement Learning." Neural Information Processing Systems (NeurIPS 2025).

Benchmarking Batch Normalization in HLS4ML Designs

Clay, D., Hauck, S.Donovan Clay, Scott Hauck (2025). "Benchmarking Batch Normalization in HLS4ML Designs." Honors Undergraduate Thesis, University of Washington.

Teaching

Machine Learning (CSE 546/446)

Digital Design (CSE 369)