Sanchit Gandhi
Research Scientist at Mistral AI
London, England, United States
Summary
🤩
Rockstar🎓
Top SchoolSanchit Gandhi is a research scientist specializing in speech and NLP with four years of hands-on experience building and leading open-source audio ML at Hugging Face and now Mistral AI. He led development of high-impact projects such as Distil-Whisper and Whisper-JAX—dramatically improving inference speed and robustness—and supervised low-latency Parler-TTS while maintaining core Audio and JAX model implementations in Transformers and Diffusers. A pragmatic engineer-researcher, he’s converted and integrated complex audio diffusion and speech models (AudioLDM, MusicLDM, MusicGen, Wav2Vec2) across PyTorch and Flax, contributing performance and usability improvements like CPU offload and adapter weights. His background from Cambridge includes interpretable deep-learning work and even leadership outside tech as president of a university sports club, reflecting an ability to coordinate teams and projects under pressure.
4 years of coding experience
3 years of employment as a software developer
Information and Computer Engineering MEng, Information and Computer Engineering MEng at University of CambridgeGCSE & A-Levels, GCSE & A-Levels at Bristol Grammar School
Stackoverflow
31reputation
1kreached
1answer
0questions
Github Skills (24)
transformers10
speech-to-text10
pytorch10
python10
wafv210
diffusion-models10
machine-learning10
text-to-speech10
diffusers10
flax10
diffusion-probabilistic-models10
diffusion-probabilistic10
deeplearning-ai10
deep-learning10
natural-language-processing9
Programming languages (9)
Github contributions (5)
huggingface/parler-tts
Feb 2024 - Jan 2025
Inference and training library for high-quality TTS models.
Role in this project:
ML Engineer
ML EngineerContributions:23 reviews, 11 PRs, 21 pushes in 10 months
Contributions summary:Sanchit's commits primarily involve the development and modification of code within the `stable_speech` directory, indicating contributions to Stability AI's text-to-speech model. Their work includes setting up the project with necessary dependencies, integrating modules from musicgen, and creating and modifying core model components. The user also worked on adding a new audio classification feature and implementing regularisation, demonstrating involvement across multiple phases of the project.
huggingface/transformers
Feb 2022 - Jan 2023
🤗 Transformers: State-of-the-art Machine Learning for Pytorch, TensorFlow, and JAX.
Role in this project:ML Engineer
ML EngineerContributions:1107 reviews, 57 commits, 320 PRs in 11 months
Contributions summary:Sanchit primarily contributed to the integration and development of machine learning models within the Hugging Face Transformers library. Their work involved converting PyTorch models to Flax, adding new components like adapter weights to the Flax Wav2Vec2 implementation, and fixing bugs within existing model implementations. Additionally, the user was involved in the development of Flax models for Speech-Encoder-Decoder, including testing and code consistency.
pythonbertspeech-recognitionstate-of-the-artflax
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.