10 AI lab post-training researchers

What these AI lab post-training researchers share is RL and post-training work on large models: three on Gemini at Google DeepMind, others on Claude at Anthropic, Llama at Meta and Amazon's foundation models, plus OpenAI, Cohere, Microsoft AI and Character.ai. A research director, an RL systems engineer and members of technical staff all appear, which suits a team choosing an RLHF approach, a reward signal or a training stack.

10 professionals7 companiesData retrieved October 9, 2026

Professionals to explore

  1. Aliaksei Severyn

    LinkedIn

    Experience: Research Director | Principal Research Scientist · Google DeepMind

    $1,683 per 30-minute call

    Aliaksei Severyn has been Research Director | Principal Research Scientist at Google DeepMind since November 2024, describing the role as Gemini post-training and user feedback lead, including Gemini app post-training and free-tier Gemini model post-training. Google research titles run from Senior Research Scientist (2016-2020) to Senior Staff Research Scientist (2023-2024). The best fit here for questions on turning user feedback into a post-training signal.

    Is this you? Claim your profile and get paid for calls.

  2. Devamanyu Hazarika

    LinkedIn

    Experience: Research Scientist · Meta

    $673 per 30-minute call

    Devamanyu Hazarika is a Research Scientist at Meta since January 2025, working on post-training in the Llama Research team at Meta Superintelligence Labs. As Senior Applied Scientist at Amazon (2024-2025), Devamanyu Hazarika was a core developer of the Amazon Nova family of models and co-led the model alignment team. A fit for questions on alignment work across two labs' model families.

    Is this you? Claim your profile and get paid for calls.

  3. Haochen Zhang

    LinkedIn

    Experience: Member of Technical Staff · Anthropic

    $808 per 30-minute call

    Working on RL and post-training of Claude models as a Member of Technical Staff at Anthropic since April 2024. Haochen Zhang was a PhD student in Computer Science at Stanford University (2019-2024), working on the theoretical foundations of modern learning algorithms, with a focus on the interpretability of large pre-trained models. A good pick when a question needs both RL practice and theory.

    Is this you? Claim your profile and get paid for calls.

  4. Haoran Li

    LinkedIn

    Experience: Former Research TL @ GenAI · Meta

    $606 per 30-minute call

    Haoran Li is Head of Post-training & Data (Research) at Character.ai since May 2025. As Research TL @ GenAI at Meta (2017-2024), Haoran Li led LLM post-training for Character Chatbots and AI Studio, covering fine-tuning, RLHF, evaluation and the user data flywheel. That record suits questions on post-training conversational products and building a data loop from user interactions.

    Is this you? Claim your profile and get paid for calls.

  5. Jiahui Li

    LinkedIn

    Experience: Senior Research Engineer, AGI Post-Training · Amazon

    $539 per 30-minute call

    Senior Research Engineer, AGI Post-Training at Amazon since April 2025, after Research Engineer, AGI Post-Training (2023-2025) and Software Engineer, Alexa AI (2020-2023). Jiahui Li builds large-scale asynchronous and off-policy RL systems for foundation model post-training, with agentic RL and distributed training named as focus areas, so questions on RL infrastructure belong here.

    Is this you? Claim your profile and get paid for calls.

  6. Jyh-Jing Hwang

    LinkedIn

    Experience: Research Scientist · Google DeepMind

    $808 per 30-minute call

    At Google DeepMind since September 2025, Jyh-Jing Hwang is a Research Scientist focused on Gemini post-training research and agentic RL. The prior role was Research Scientist & TLM at Waymo (2020-2025), with research interests across AI, ML, computer vision and robotics. A fit for questions on applying RL to agent behavior.

    Is this you? Claim your profile and get paid for calls.

  7. Rohan Kshirsagar

    LinkedIn

    Experience: Member of Technical Staff · OpenAI

    $808 per 30-minute call

    Member of Technical Staff at OpenAI since March 2024 in post-training research, listing work on offline and online predictiveness and evals, diversity collapse, hack robustness, RL interpretability and user signal reward modeling. Rohan Kshirsagar was Staff Machine Learning Engineer at Airbnb (2018-2022) and Senior Software Engineer in the NLP Group at Bloomberg (2013-2016). Reward hacking and evaluation questions fit well.

    Is this you? Claim your profile and get paid for calls.

  8. Shunyu Yao

    LinkedIn

    Experience: Senior Staff Research Scientist · Google DeepMind

    $1,346 per 30-minute call

    Shunyu Yao has been a Senior Staff Research Scientist at Google DeepMind since September 2025 on Gemini post-training research, co-leading a project on research automation, and was Member of Technical Staff at Anthropic (2024-2025), a research scientist on reinforcement learning who helped build the RL fundamentals team and contributed to Claude 3.7 Sonnet. Comparing RL practice across Anthropic and Google DeepMind is the natural question here.

    Is this you? Claim your profile and get paid for calls.

  9. Yannis Flet-Berliac

    LinkedIn

    Experience: Research Scientist · Cohere

    $606 per 30-minute call

    Yannis Flet-Berliac has been a Research Scientist at Cohere since April 2024, focusing on LLM post-training and reinforcement learning research. Before Cohere came Research Scientist at InstaDeep (2023-2024) and a postdoctoral role at Stanford University (2022-2023) working on offline reinforcement learning, which makes them a fit for questions on offline RL methods for language models.

    Is this you? Claim your profile and get paid for calls.

  10. Yunzhe Tao

    LinkedIn

    Experience: Member of Technical Staff · Microsoft AI

    $606 per 30-minute call

    Post-training on the superintelligence team as Member of Technical Staff at Microsoft AI since May 2026. At Anuttacon, where Yunzhe Tao was Founding Research Scientist (2024-2026), they led post-training of humanized AI and model behavior and post-training of LLMs for code generation; Staff Research Scientist, Tech Lead at ByteDance (2021-2024) came before. Model behavior and code generation post-training are the strongest areas.

    Is this you? Claim your profile and get paid for calls.

Choose the right perspective

For RL infrastructure and scaling, Jiahui Li builds asynchronous and off-policy RL systems at Amazon. Reward modeling from user signals points to Rohan Kshirsagar and Aliaksei Severyn, consumer chatbot post-training to Haoran Li, and agentic RL to Jyh-Jing Hwang; Shunyu Yao and Haochen Zhang cover RL work at Anthropic, with Shunyu Yao now on Gemini.

Questions to take into the conversation

  1. 01When user feedback from a consumer chat app feeds a reward model, how do you keep the model from learning to game that signal?
  2. 02At what model size does an off-policy or asynchronous RL setup start to pay for its extra complexity over a simpler fine-tuning loop?
  3. 03How do you decide whether an agentic coding or tool-use behavior should be taught through RL or through supervised data?

About this directory

This is a professional research starting point based on business profile data retrieved on . Titles and companies reflect that source snapshot; roles marked former had ended by then. Check the linked profiles for current details. Inclusion does not imply Instant Expert membership or availability.

Other perspectives in Artificial intelligence