AI Scientist - Audio

About Mistral

At Mistral, we are on a mission to democratize AI, producing frontier intelligence for everyone, developed in the open, and built by engineers all over the world. We are a dynamic, collaborative team passionate about AI and its potential to transform society. Our diverse workforce thrives in competitive environments and is committed to driving innovation, with teams distributed between Europe, the USA, and Asia. We are creative, low-ego, and team-spirited.

At Mistral, we develop models for the enterprise and for consumers, focusing on delivering systems which can really change the way in which businesses operate and which can integrate into our daily lives. All while releasing frontier models open-source, for everyone to try and benefit.

Mistral is hiring experts in the training of large language models and distributed systems. Join us to be part of a pioneering company shaping the future of AI.

What will you do

• Research and develop novel methods to push the frontier of large language models
• Work across use cases (e.g., reasoning, code, agents) and modalities (e.g., text, image, and speech)
• Build tooling and infrastructure to allow training, evaluation, and analysis of AI models at scale
• Work cross-functionally with other scientists, engineers, and product teams to ship AI systems which have a real-world impact

About you

• An expert in speech input/output methodologies (specific to audio)
• You are a highly proficient software engineer in at least one programming language (Python or other, e.g., Rust, Go, Java)
• You have hands-on experience with AI frameworks (e.g., PyTorch, JAX) or distributed systems (e.g., Ray, Kubernetes)
• You have high engineering competence. This means being able to design complex software and make it usable in production
• You are a self-starter, autonomous, and a team player

It would be ideal if

• You have experience working with large-scale speech-language models
• You have hands-on experience with training large transformer models in a distributed fashion
• You can navigate the full MLOps stack, for instance, fine-tuning, evaluation, and deployment
• You have a strong publication record in a relevant scientific domain