About the role#
Crusoe's Managed AI team develops services that allow customers to customize and run large language models without managing the underlying infrastructure. Our platform covers post-training, where customers adapt models using supervised fine-tuning and reinforcement learning, and managed inference, where we run models on our GPU fleet. As an AI Applied Research PhD Intern, you will own a research question spanning from model training to serving. You will run experiments on frontier open-weight models using large GPU clusters, turn your findings into platform features, and present your results to our team and leadership. This is a 12-week, full-time, in-person internship.
What you'll need#
- Current enrollment in a PhD program in Computer Science, Machine Learning, Electrical Engineering, or a related field, with an expected graduation date between Fall 2027 and Spring 2029.
- At least 2 years of research experience in machine learning or ML systems.
- Hands-on experience with large language models, including training, fine-tuning (SFT, RLHF, DPO, GRPO/PPO), or inference optimization.
- Proficiency in Python and PyTorch, with the ability to read and modify large ML codebases.
- Experience designing controlled experiments, selecting appropriate metrics, and distinguishing real effects from noise.
- Knowledge of how LLM training and inference run on GPUs, including batching, memory, mixed precision, and parallelism.
- Strong communication skills to explain complex results in writing and in person.
- Bonus qualifications include publications at venues such as NeurIPS, ICML, ICLR, MLSys, ACL, EMNLP, or COLM; experience with inference engines like vLLM or TensorRT-LLM; training frameworks such as Megatron-LM or DeepSpeed; and skills in quantization, GPU kernels (CUDA, Triton), or building agent environments.
What you'll do#
- Own a research question that bridges the gap between model training and inference.
- Run experiments on frontier open-weight models using large GPU clusters.
- Develop recipes and platform features based on your research findings.
- Present your results to the team and leadership.
- Collaborate with engineers to transition research into production.
Location & details#
- Location: San Francisco, California, USA.
- Term: Summer 2027 (12 weeks).
- Work Modality: On-site.
- Compensation: $80 per hour.
- Benefits: Includes a one-time housing stipend, access to HealthiestYou and Calm, paid holiday and volunteer days, and a commuter allowance.


