Intern - Frontier Research

Please submit your CV in English.
This is a paid internship.

About the role:

At Owkin, we believe the future of healthcare R&D lies at the intersection of science and AI technologies. As a Research Intern in our Data Science team, you will play a pivotal role in shaping and showcasing this future. You will help develop new cutting-edge AI agent technologies applied to biotech challenges by fine-tuning LLMs. Your goal will be to create high quality training data and fine-tune LLMs on a multi-GPU cluster in order to increase its computational biology capabilities.

In this research internship you will likely face several challenges inherent to LLM fine-tuning (such as catastrophic forgetting) and propose novel ideas to mitigate them. You will have autonomy and the opportunity to have a high impact on your project.

You will be part of the Frontier Research team which works on long-term research projects at Owkin. You will be an active member of the team and present research articles on a regular basis. Working closely with other research scientists you will have the opportunity to participate in the creation of a biomedical AI agent.

In particular, you will: 

Core Duties

• Perform fine-tuning of LLMs applied to better understand complex biological modalities
• Create and curate high-quality training data
• Participate in the creation of internal benchmarks and evaluate your models on these benchmarks
• Write and submit a research paper
• Document what you learn and build along the way

Other duties

• Integrate fine-tuned models into a dedicated agent and evaluate its performances
• Benchmark existing and competing agents
• Stay up-to-date with the latest literature on LLMs and agents
• Present research papers to the Data science group

About you

Required qualifications / experience:

• Currently pursuing a Master’s degree in Computer Science, AI, mathematics, data science, or related field
• Good programming skills in Python  
• Some experience with AI/ML frameworks (Pytorch is preferred)
• Knowledge in machine learning and deep learning methods (e.g. linear models, transformer architecture, gradient descent, vision-language models, …)
• Familiarity with large language models (LLMs). Experience with Anthropic or OpenAI models is a plus
• Basic understanding of AI agents and prompt engineering
• Ability to pick up new technologies quickly
• Good communication skills – can communicate complex scientific topics to a general audience
• Self-motivated and works well independently or in teams

Preferred qualifications/bonus:

• Experience with HuggingFace libraries (transformers, trl, datasets, …)
• Experience working with cloud super-computers (slurm, …)
• Interest in life sciences, biotechnology, or pharmaceuticals
• Experience creating interactive demos or visualizations for non-technical audiences
• Previous projects involving AI-powered tools or applications
• Familiarity with research in machine learning
• Experience tinkering with AI agent frameworks (like LangGraph or smolagents)
• Contributions to open-source projects
  
  Internal use: #LI-CJ1