10 Mistral AI research scientists
Pre-training, reinforcement learning post-training, reasoning and multimodal models are the areas seven of these Mistral AI research scientists name for their own work, and all ten hold Mistral roles now, one as Research Scientist Lead. Seven are based in California, the others in Paris, Warsaw and Atlanta; for questions on choosing training and evaluation priorities, their earlier roles at Google DeepMind, NVIDIA, Apple, Meta and Ai2 give points of comparison.
Professionals to explore
Andy Ehrenberg
LinkedInExperience: AI Research Scientist · Mistral
$1,683 per 30-minute call
Two AssemblyAI roles precede Andy Ehrenberg's current AI Research Scientist post at Mistral (since 2024): Senior Researcher (2024) and Research Engineer (2022-2024). Ehrenberg was also a Data Scientist at C3 AI (2021-2022) and a Research Intern at UCL DARK Lab (2020-2021), and is based in the San Francisco Bay Area. Questions comparing research work at AssemblyAI and Mistral suit this record.
Is this you? Claim your profile and get paid for calls.
Jan Ludziejewski
LinkedInExperience: AI Research Scientist · Mistral
$1,683 per 30-minute call
Based in the Warsaw area, Jan Ludziejewski has been AI Research Scientist at Mistral since 2025, working on pretraining, and is a PhD Candidate in Machine Learning at the University of Warsaw. Earlier roles were Badacz-doktorant at IDEAS NCBR (2022-2025) and Machine Learning Specialist at QED Software (2017-2022). Pretraining questions, and how doctoral research carries into pretraining work at Mistral, suit Ludziejewski.
Is this you? Claim your profile and get paid for calls.
Khyathi Raghavi Chandu
LinkedInExperience: AI Research Scientist · Mistral
$1,683 per 30-minute call
Pre-training, post-training and data make up the span of Khyathi Raghavi Chandu's work as AI Research Scientist at Mistral since 2024, with multimodal LLMs and robotics as stated interests. Earlier roles were Research Scientist at Ai2 (2023-2024), Research Scientist (2022-2023) and Research Intern (2021) at Meta, and Graduate Research Assistant at Carnegie Mellon University (2016-2021). Training data questions for multimodal models fit Chandu.
Is this you? Claim your profile and get paid for calls.
Pavankumar Reddy M
LinkedInExperience: AI Research Scientist · Mistral
$1,683 per 30-minute call
Described as Multimodal Audio Lead, Pavankumar Reddy M has been AI Research Scientist at Mistral since 2024, working on pre-training and post-training alignment of language and multimodal models. Before that came four Google titles: Staff Software Engineer / Manager - Gemini (2023-2024), Staff Software Engineer / Manager - Assistant NLP (2022-2023), Senior Software Engineer / Manager - Assistant NLP (2021-2022) and Senior Software Engineer - Assistant NLP (2020-2021). Audio model questions fit.
Is this you? Claim your profile and get paid for calls.
Ram Ramrakhya
LinkedInExperience: Research Scientist · Mistral
$1,683 per 30-minute call
RL post-training research for reasoning and agents has been Ram Ramrakhya's work as a Research Scientist at Mistral since 2026, after Graduate Research Assistant at Georgia Institute of Technology (2021-2025), where Ramrakhya did a CS PhD. Internships were Research Intern at Apple (2025), Research Scientist Intern at Meta (2024) and Research Intern at Ai2 (2023). Based in Atlanta, Ramrakhya fits questions on post-training for reasoning and agents.
Is this you? Claim your profile and get paid for calls.
Romain Sauvestre
LinkedInExperience: Former Research Scientist Lead · Mistral
$2,019 per 30-minute call
In 2026, Romain Sauvestre became Research Scientist Lead at Mistral after Research Scientist there (2024-2026), and is based in Paris. Earlier roles were Research Engineer at Meta (2022-2024), Engineering Manager, Machine Learning at Hyperscience (2020-2022) and Research Engineer at Imagen Technologies (2018-2020). The lead title and the earlier engineering manager role make Sauvestre the fit for questions on organizing a model research team.
Is this you? Claim your profile and get paid for calls.
Sandeep Subramanian
LinkedInExperience: Research Scientist · Mistral
$1,683 per 30-minute call
Sandeep Subramanian has been a Research Scientist at Mistral since 2024, directly after three NVIDIA titles: Research Manager (2023-2024), Senior Research Scientist (2021-2023) and Research Scientist (2020-2021). Subramanian was a PHD Student at Mila - Quebec Artificial Intelligence Institute (2016-2021) and is based in San Francisco. The NVIDIA research manager role makes Subramanian a fit for questions on managing research scientists.
Is this you? Claim your profile and get paid for calls.
Srijan Mishra
LinkedInExperience: AI Research Scientist · Mistral
$1,683 per 30-minute call
Multimodal LLMs and reinforcement learning are Srijan Mishra's focus as AI Research Scientist at Mistral since 2024. Mishra was a Machine Learning Scientist at Apple (2020-2024), a Machine Learning Engineer at Uber (2019-2020) and a Machine Learning Engineering Intern at Google (2018), and is based in the San Francisco Bay Area. Questions on combining multimodal models with reinforcement learning fit this record.
Is this you? Claim your profile and get paid for calls.
Zhenlin Xu
LinkedInExperience: AI Research Scientist · Mistral
$1,683 per 30-minute call
LLM pre- and post-training, reasoning and multimodal models are Zhenlin Xu's work as AI Research Scientist at Mistral since 2025, with recent work on aligning LLMs using synthetic data and training agentic models with reinforcement learning for tool use. Earlier roles were Machine Learning Scientist at Boson AI (2024-2025) and Applied Scientist at Amazon (2022-2024), plus research internships at Google (2021) and NVIDIA (2020). Synthetic data for alignment and reinforcement learning for tool use are the topics for Xu.
Is this you? Claim your profile and get paid for calls.
Choose the right perspective
Pre-training questions go to Jan Ludziejewski or Khyathi Raghavi Chandu; reinforcement learning, agents and reasoning to Ram Ramrakhya, Zhenlin Xu or Abhinav Rastogi; multimodal and audio work to Pavankumar Reddy M or Srijan Mishra. Romain Sauvestre (Research Scientist Lead) and Sandeep Subramanian (formerly a Research Manager at NVIDIA) fit questions on running a research team.
Questions to take into the conversation
- 01How does a research team decide how much compute goes to pre-training versus post-training experiments?
- 02What evaluation signals show that reinforcement learning post-training is improving reasoning rather than fitting a benchmark?
- 03How do you assemble training data for an audio or multimodal model when labeled data is scarce?
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.