AI Applied Research PhD Intern, Summer 2027

Crusoe•Sunnyvale, CA
•Onsite

About The Position

Crusoe’s Managed AI team builds services that help customers customize and run large language models without managing infrastructure. Its post-training platform supports supervised fine-tuning and reinforcement learning, while its managed inference service serves open-weight and fine-tuned models, including LoRA adapters, on Crusoe’s GPU fleet. Training and serving are connected: models trained on the platform can be deployed on the same infrastructure, and reinforcement learning training relies on high-throughput inference for rollouts. The AI Applied Research PhD Intern will own a research question spanning training and serving, such as improving the reliability of reinforcement learning on customer tasks or making inference faster and more cost-efficient without sacrificing quality. The intern will run experiments with frontier open-weight models on large GPU clusters, turn findings into recipes and platform features for customers, and present results to the team and leadership. This is a 12-week, full-time, in-person internship for candidates interested in putting research into production. The internship program pairs high-impact work with one-on-one mentorship and direct access to organizational leaders. Interns gain hands-on experience, take on real responsibility, and receive support for professional growth and building connections.

Requirements

  • Currently enrolled in a PhD program in Computer Science, Machine Learning, Electrical Engineering, or a related field, with expected graduation between Fall 2027 and Spring 2029.
  • At least 2 years of machine learning or ML systems research experience, including hands-on experience with large language models through training or fine-tuning (such as SFT, RLHF, DPO, GRPO/PPO) or inference optimization.
  • Strong Python and PyTorch skills, and comfort reading and modifying large ML codebases.
  • Ability to design controlled experiments, select suitable baselines and metrics, and distinguish real effects from noise.
  • Working knowledge of how LLM training and inference run on GPUs, including batching, memory, mixed precision, and parallelism.
  • Ability to explain complex results clearly in writing and in person, and to work well with engineers translating research into products.

Nice To Haves

  • Publications at venues such as NeurIPS, ICML, ICLR, MLSys, ACL, EMNLP, or COLM on RL for LLMs, reward modeling, agents, efficient training, or efficient inference.
  • Hands-on experience with inference engines such as vLLM, SGLang, or TensorRT-LLM, or training frameworks such as verl, OpenRLHF, TRL, Megatron-LM, or DeepSpeed.
  • Experience with quantization, speculative decoding, GPU kernels (CUDA or Triton), or LoRA/PEFT methods.
  • Experience building agent environments, tool-use sandboxes, or evaluation harnesses.
  • Open-source contributions to training or inference libraries.

Responsibilities

  • Own a research question related to the path from LLM training to serving.
  • Run experiments on frontier open-weight models using large GPU clusters.
  • Develop research findings into recipes and platform features for customers.
  • Present results to the team and leadership.

Benefits

  • Compensation of $80 per hour.
  • One-time housing stipend.
  • Access to HealthiestYou and Calm.
  • Paid holiday and volunteer days.
  • Commuter allowance.
© 2026 Teal Labs, Inc
Privacy PolicyTerms of Service