Research Intern, Agent RL Training

The role

You will work with a mentor to explore applying large language models to content understanding, recommendation, agentic web browsing, and autonomous multi-step task completion. You will drive experiments, propose ideas, and iterate on LLM post-training and agent capabilities.

What you will do

• Collaborate with a full-time mentor to identify research directions for applying LLMs to products
• Independently run end-to-end SFT experiments on LLM-based agents
• Assist with RL-related exploration such as reward design and training iteration
• Curate and build high-quality training datasets including instruction-following, preference pairs, agent trajectories, and synthetic data
• Contribute to public publications

Required qualifications

• Highly motivated and committed
• Genuine passion for research
• Independently capable of end-to-end model SFT
• Basic understanding of RL-based post-training methods (RLHF, DPO, PPO, GRPO, etc.)
• Excellent taste in model behavior
• Strong Python and PyTorch skills

Nice to have

• Publication at a top-tier venue (NeurIPS, ICML, ICLR, ACL, EMNLP, or equivalent)
• Experience with multi-node distributed training (FSDP, DeepSpeed, Megatron-LM)
• Proficiency in writing custom GPU kernels with Triton or CUDA
• Experience building synthetic data pipelines for agent training
• Familiarity with open-source RL frameworks: TRL, OpenRLHF, veRL/vLLM

About the company

NewsBreak is a local life and news platform with more than 40 million monthly users. It includes NewsBreak, Nearby AI, and NDot AI.