ML Research Engineer

At Hugging Face, we’re on a journey to democratize good AI. We are building the fastest growing platform for AI builders with over 5 million users & 100k organizations who collectively shared over 1M models, 300k datasets & 300k apps. Our open-source libraries have more than 400k+ stars on Github.

About the Role

2025 will be the year of Agents.

LLM agents, what are these? In short, agentic systems are a way to connect an LLM to the outside world, to let it perform actions and observations. The promises of such a setup are immense : it enables an LLM to get external knowledge to ground its reasoning, use reliable tools for specific tasks, and even self-correct errors through additional iterations.

An agentic system requires a powerful enough LLM to reason and plan actions. Now with the progress of LLMs in general and specifically open models overtaking closed ones, we have all weapons to build great agents. So Hugging Face is...

At Hugging Face, we’re on a journey to democratize good AI. We are building the fastest growing platform for AI builders with over 5 million users & 100k organizations who collectively shared over 1M models, 300k datasets & 300k apps. Our open-source libraries have more than 400k+ stars on Github.

About the Role

LLM Agents are an exciting new field giving LLMs more autonomy and tools to solve more complex and longer tasks. Building capable agents is both exciting and challenging as it lies at the intersection of post-training, inference, orchestration as well as UX design.

During this internship you will work specifically on agents that can reason over data using code. This work will include the whole agent stack from preparing fine-tuning datasets, training recipes and inference orchestration. You will leverage our compute CPU and H100 clusters for large scale processing and model training. Finally we’ll release the full recipe as...

At Hugging Face, we’re on a journey to democratize good AI. We are building the fastest growing platform for AI builders with over 5 million users & 100k organizations who collectively shared over 1M models, 300k datasets & 300k apps. Our open-source libraries have more than 400k+ stars on Github.

About the Role

LLM Agents are an exciting new field giving LLMs more autonomy and tools to solve more complex and longer tasks. Building capable agents is both exciting and challenging as it lies at the intersection of post-training, inference, orchestration as well as UX design.

During this internship you will work specifically on agents that can reason over data using code. This work will include the whole agent stack from preparing fine-tuning datasets, training recipes and inference orchestration. You will leverage our compute CPU and H100 clusters for large scale processing and model training. Finally we’ll release the full recipe as...

At Hugging Face, we’re on a journey to democratize good AI. We are building the fastest growing platform for AI builders with over 5 million users & 100k organizations who collectively shared over 1M models, 300k datasets & 300k apps. Our open-source libraries have more than 400k+ stars on Github.

About the Role

At Hugging Face, we're dedicated to democratizing machine learning and making cutting-egg models accessible to everyone. We focus on developing open-source tools and models that push the boundaries of AI while remaining efficient and user-friendly.

Aligned with this, we've recently released SmolVLM [1], a state-of-the-art, fully open-source VLM that's small, fast, and memory-efficient. SmolVLM stands out for its ability to run on limited computational resources, making it deployable on local setups like laptops and edge devices. This opens up new possibilities for reducing inference costs and enabling user...

At Hugging Face, we’re on a journey to democratize good AI. We are building the fastest growing platform for AI builders with over 5 million users & 100k organizations who collectively shared over 1M models, 300k datasets & 300k apps. Our open-source libraries have more than 400k+ stars on Github.

About the Role

Post-training is an exciting and fast-moving field of research that is used to enhance the performance of large language models and enable them to follow human instructions. The post-training team at Hugging Face is pushing the frontier of model capabilities by developing recipes [1] that produce state-of-the-art models like Zephyr [2] and NuminaMath [3], which won the 1st Progress Prize of the AI Math Olympiad.

During this internship, you will work alongside the post-training team to implement cutting-edge research and make it accessible to the global AI community in the form of code, datasets, and models. Topics include...

At Hugging Face, we’re on a journey to democratize good AI. We are building the fastest growing platform for AI builders with over 5 million users & 100k organizations who collectively shared over 1M models, 300k datasets & 300k apps. Our open-source libraries have more than 400k+ stars on Github.

About the Role

High-quality datasets are the foundation of strong LLMs, yet, most labs releasing state-of-the-art models are vague when it comes to the pretraining data. At Hugging Face we want to enable all the community to build the best models by building and open-sourcing the finest datasets. FineWeb and FineWeb-Edu are examples of very strong, web-scale datasets we released this year while also open-sourcing the distributed processing library datatrove.

During this internship you will work alongside the FineWeb team and build the next generation of high-quality web data, by running distributed data processing and ablating the data...

At Hugging Face, we’re on a journey to democratize good AI. We are building the fastest growing platform for AI builders with over 5 million users & 100k organizations who collectively shared over 1M models, 300k datasets & 300k apps. Our open-source libraries have more than 400k+ stars on Github.

About the Role

Smol models are an exciting area of research as they enable cheaper inference and can be run on-device allowing for more customization and ensuring privacy. The SmolLM team at Hugging Face is pushing the frontier of smol models by building high quality pre-training and post-training datasets [1,2], and applying the latest architecture and training techniques to develop state-of-the-art models [2,3]. The dataset processing can leverage our scalable CPU cluster and the models are trained on a state-of-the-art H100 cluster with close to 100 nodes.

In this internship you will work alongside the SmolLM team and work towards...

About the Role

Are you a PyTorch aficionado with a knack for translating deep learning theory into practical PyTorch implementations? We're seeking a PyTorch expert to join us at the forefront of computer vision research and development. As part of our founding team, you'll dive headfirst into cutting-edge deep learning techniques, shaping the future of computer vision foundation models.

What You'll Do

• Translate complex deep learning concepts into efficient PyTorch code, pushing the boundaries of computer vision.
• Dive into the forefront of deep learning research, tackling daring and technically complex challenges.
• Work closely with the founding team, contributing to the development of groundbreaking solutions.
• Embrace a fast-paced, dynamic environment where innovation thrives and new ideas are welcomed.
• If you have an ear for music, consider it a bonus – we're building something...

About the Role

Are you a PyTorch aficionado with a knack for translating deep learning theory into practical PyTorch implementations? We're seeking a PyTorch expert to join us at the forefront of computer vision research and development. As part of our founding team, you'll dive headfirst into cutting-edge deep learning techniques, shaping the future of computer vision foundation models.

What You'll Do

• Translate complex deep learning concepts into efficient PyTorch code, pushing the boundaries of computer vision.
• Dive into the forefront of deep learning research, tackling daring and technically complex challenges.
• Work closely with the founding team, contributing to the development of groundbreaking solutions.
• Embrace a fast-paced, dynamic environment where innovation thrives and new ideas are welcomed.
• If you have an ear for music, consider it a bonus – we're building something...