Research Intern, Agent RL Training
Core
Research Intern developing LLM-based agents for NewsBreak's core business (content understanding, recommendation, agentic web browsing, autonomous multi-step task completion) using large language models.
Role type
Research Intern (LLM Post-Training & Agent RL)
Builds
LLM-based agents for local life and news platform
Domain
AI/ML, Large Language Models, Reinforcement Learning
Deliverable
production ML models
Required skills
Python, PyTorch, RLHF, DPO, PPO, GRPO, end-to-end model SFT, reward design, training iteration, dataset curation, synthetic data generation
Preferred skills
Top-tier publication experience, multi-node distributed training (FSDP, DeepSpeed, Megatron-LM), custom GPU kernels (Triton, CUDA), open-source RL frameworks (TRL, OpenRLHF, veRL/vLLM)
Technologies
PyTorch, Triton, CUDA, FSDP, DeepSpeed, Megatron-LM, TRL, OpenRLHF, veRL, vLLM
Responsibilities
Identify high-impact research directions for LLM applications, run end-to-end SFT experiments, assist with RL exploration (reward design, training iteration), curate and build training datasets, contribute to public publications
Seniority
Intern