Devendra Singh Chaplot, AI Research Scientist at Mistral AI
Background and Education
Devendra Singh Chaplot holds a Ph.D. in Machine Learning from Carnegie Mellon University (CMU), where he focused on building intelligent autonomous navigation agents. He studied under prominent professors including Ruslan Salakhutdinov, Abhinav Gupta, and Saurabh Gupta. These collaborations highlight his formidable background in Computer Vision and Robotics. Prior to his doctoral studies, Chaplot received his Bachelor's degree in Computer Science from IIT Bombay, with a minor in Applied Statistics.
Professional Experience
Mistral AI (August 2023 - Present):
- AI Research Scientist: Chaplot is currently building next-generation AI models at Mistral AI. His responsibilities include the development of large language models and implementing advanced AI research to enhance Mistral’s product offerings.
Facebook AI Research:
- Research Scientist: Worked at the intersection of Machine Learning, Computer Vision, and Robotics. His tenure at Facebook AI Research (FAIR) included significant contributions such as winning the CVPR Habitat ObjectNav Challenge and producing research that received Best Paper and Best Demo awards at leading AI conferences.
Samsung Electronics HQ, South Korea:
- Software Engineer: Briefly worked at Samsung Electronics HQ before embarking on his Ph.D.
Key Achievements and Contributions
- Series B Funding Announcement: Mistral AI raised $640M in Series B funding, reaching a valuation of over $6 billion as of June 2024. This funding showcases the company’s rapid growth under Chaplot’s contributions.
- Mistral AI Fine-tuning API: Announced in June 2024, the API offers serverless fine-tuning using LoRA adapters, which is positioned as both fast and cost-efficient.
- Mistral 8x7B Release: In December 2023, Chaplot announced the release of Mistral 8x7B, a high-quality sparse mixture of experts model under the Apache 2.0 license.
- Outstanding Model Performance:
- Mistral Large: Achieved 81.2% MMLU, outperforming other cutting-edge models like Claude 2, Gemini 1.0 Pro, and GPT-4.
- Mistral-NeMo 12B (July 2024): Developed in collaboration with NVIDIA, this model outperformed Gemma2 9B and Llama3 8B, emphasizing the organization’s commitment to superior AI performance.
Publications and Research Highlights
Chaplot has a prolific portfolio of research, often featured in high-impact journals and conferences. Some of his notable projects include:
- SEAL: Self-supervised Embodied Active Learning: Published at NeurIPS 2021, this research presents advancements in self-supervised learning within embodied AI.
- Goal-Oriented Semantic Exploration: A NeurIPS 2020 paper focusing on object goal navigation.
- Learning to Explore using Active Neural SLAM: Presented at ICLR 2020, this project illustrates novel methods in active learning and robotic exploration.
Public Recognition and Media Features
Chaplot’s research successes and contributions to AI have been recognized widely, including features in renowned media outlets such as TechCrunch, Engadget, and MIT Technology Review. He has delivered numerous talks at prestigious conferences, reinforcing his position as a thought leader in AI.
Current Position and Role at Mistral AI
In his current capacity at Mistral AI, Chaplot is significantly driving forward the development and deployment of large-scale AI models that stand out in performance on industry benchmarks. His consistent track record of innovation and excellence is pivotal to Mistral AI’s strategic vision and market positioning.
Contact Information:
- Email: [dc@mistral.ai](mailto:dc@mistral.ai)
- LinkedIn: [Devendra Chaplot](https://www.linkedin.com/in/devendra-chaplot-b1605534)
This detailed profile encapsulates Devendra Singh Chaplot’s professional journey, highlighting his educational background, work experience, strategic contributions to Mistral AI, and his recognized research within the broader AI community.
