Intern, AI Engineering

Responsibilities

• We are seeking exceptional graduate students to join our AI Lab as Research Interns in San Francisco.
• You'll work on fundamental problems in LLM-based agentic systems and efficient AI infrastructure, with opportunities to publish your research while making direct impact on production systems serving enterprise customers.
• Research Areas:
• LLM Agent Systems: Design and implement intelligent agent architectures for complex enterprise automation tasks, including multi-agent collaboration, MCP, and reasoning frameworks
• Efficient LLM Fine-tuning: Develop novel methods for parameter-efficient adaptation, alignment, and reinforcement learning for large language models
• High-Performance LLM Inference: Optimize inference pipelines through systems-level innovations, kernel development, and deployment strategies
• In This Role, You Will Also Be Responsible To:
• Conduct original research on LLM agent architectures and optimization techniques
• Develop and evaluate novel algorithms with both academic rigor and production feasibility
• Present your work at internal research seminars and external conferences
• Mentor and collaborate with LLM engineers on implementation and deployment

Requirements

• Currently pursuing MS/PhD in Computer Science, Machine Learning, Natural Language Processing, or related fields
• Publishations at top-tier venues (ICML, NeurIPS, ICLR, ACL, EMNLP, NAACL)
• Strong programming skills in Python and PyTorch
• Ability to work in-person at our San Francisco office
• Ability to work independently and collaborate across research and engineering teams
• Preferred Experience with self-evolving agent systems
• Proficiency in CUDA programming and custom kernel development for LLM operations
• Background in reinforcement learning-based LLM fine-tuning
• Track record of contributions to production inference systems such as vLLM, TensorRT-LLM, SGLang, or Hugging Face ecosystem
• Experience bridging academic research with production systems
• Open-source contributions to widely-used ML infrastructure projects