Research Intern – Reinforcement Learning (RL) -… at Level AI · San…

Summary: Level AI is a fast-growing Series C AI company focused on building Agentic AI systems for customer experience. They are seeking a Research Intern in Reinforcement Learning to design and build RL environments, develop RL agents, and collaborate with engineering teams to deploy learning systems at scale. Responsibilities: - Design and build reinforcement learning environments that model real-world customer interaction workflows - Design RL agents that learn from these environments using real-world interaction data, rewards, and feedback loops - Define reward models and feedback loops using real-world signals (outcomes and human feedback) - Enable learning from production data by structuring interaction traces into training-ready datasets for offline and online learning - Experiment with multi-agent systems and simulation frameworks for complex coordination and decision-making - Collaborate with engineering and product teams to deploy, evaluate, and iterate on learning systems in production at scale Required Qualifications: - Currently pursuing (or recently completed) a degree in Computer Science, AI, Machine Learning, or related field - Strong understanding of reinforcement learning fundamentals - Familiarity with RL environments and training libraries such as Verl and Tinker - Strong foundation in probability, math, and optimization - Passion for building real-world AI systems Preferred Qualifications: - Experience with RLHF, LLM/SLM fine-tuning, or model alignment - Exposure to agent-based systems or multi-agent RL - Prior research, projects, or publications in RL or applied ML - Experience working with large-scale or production datasets Required Skills: Reinforcement Learning, RL environments, Training libraries - Verl, Training libraries - Tinker, Probability, Mathematics, Optimization, RLHF, LLM fine-tuning, SLM fine-tuning, Model alignment, Agent-based systems, Multi-agent Reinforcement Learning, Applied Machine Learning, Large-scale datasets, Production datasets