SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Normal Computing is seeking an AI Research Engineer to advance agentic LLMs and reinforcement learning for code generation. This is a research-focused role where you'll design and implement multi-agent and RL approaches, build research prototypes that integrate with production systems, and create rigorous evaluation suites. You'll acquire and curate datasets from technical documents (PDFs, logs, tables), generate synthetic data where needed, and maintain data cards and licensing documentation. Your work will involve analyzing experiments with disciplined ablations, documenting results and decisions, and staying current on LLM agents, offline/online RL, RLHF/RLAIF, constrained decoding, and program synthesis.
The ideal candidate holds a PhD in CS, AI, or ML (or equivalent research experience) with publications in multi-agent RL, agentic AI, or RL for language/code. You should have strong Python and ML framework experience (PyTorch preferred; JAX and Hugging Face a plus), demonstrated ability to turn research into working systems with a reproducibility mindset (tests, seeds, configs, logging), and experience designing evaluation harnesses and success metrics for sequential/agentic tasks. You're comfortable acquiring and curating data from documents and logs, with good instincts about data quality and licensing. Clear communication and partnership with engineers are essential.
Bonus qualifications include research on program synthesis, codegen, constrained decoding, or execution-based rewards; experience with offline RL from tool traces or human corrections; open-source contributions (CleanRL, RLlib, AutoGen, LangGraph, CrewAI, Transformers); familiarity with semiconductor/chip domains or complex technical specs; and a track record of shipping research to production and measuring impact. This is an individual contributor role where impact comes through research and building, not management.
