Reinforcement Learning Researcher (12‑month contract)

About the role

Huawei Canada's Noah’s Ark Lab is seeking a Reinforcement Learning Researcher on a 12‑month contract to advance large language models through continual learning, agentic self‑improvement, and interactive evaluation.

Key responsibilities

• Implement and evaluate new training and evaluation pipelines for reasoning‑enhanced LLMs and tool‑using agents.
• Scale experiments on large GPU clusters and contribute scientific insights and publications.
• Design multi‑turn, verifiable reasoning tasks and benchmark agentic performance.
• Develop LLM post‑training paradigms such as RLHF, GRPO, reward‑free methods, and agentic reinforcement learning for tool‑using and browsing‑based agents.

Required profile

• PhD in Computer Science or related field, or a master’s degree with comparable experience.
• Strong foundation in deep learning, especially Transformer architectures and large‑model optimization.
• Research experience in reinforcement learning, self‑supervised learning, or language model fine‑tuning.
• At least one first‑author paper in top‑tier venues (NeurIPS, ICML, ICLR, CVPR, etc.).
• Proficiency in Python and experience with PyTorch, DeepSpeed, Megatron, and other distributed training frameworks.
• Excellent communication and writing skills with enthusiasm for open research.

Required skills

• Python
• PyTorch
• DeepSpeed
• Megatron
• Transformer architectures
• Reinforcement Learning (including multi‑agent RL)
• LLM post‑training pipelines (RLHF, GRPO/PPO, SFT, LoRA, MoE)
• Tool‑use / browsing / coding agents
• Distributed training on GPU clusters