- Company Name
- Understanding Recruitment
- Job Title
- Research Scientist (Gen AI)
- Job Description
-
Job Title: Research Scientist (Gen AI)
Role Summary: Lead deep research on large language model (LLM)-driven autonomous agents that reason, plan, and act in real‑world settings. Develop and evaluate LLM reasoning frameworks, structured tool‑use, and safe policy generation for long‑horizon decision making.
Expectations: Produce publishable research, prototype and test models on engineering and scientific tasks, collaborate cross‑functionally with systems, simulation, and infrastructure teams, and iterate rapidly in a high‑velocity R&D environment.
Key Responsibilities:
- Design, implement, and benchmark LLM reasoning and planning architectures for autonomous agents.
- Develop structured tool‑use, memory, reflection, and multi‑step workflow capabilities.
- Create and validate robust, safe policies for real‑world autonomous systems.
- Train and evaluate models using techniques such as supervised fine‑tuning (SFT), reinforcement learning from human feedback (RLHF), direct policy optimization (DPO), or verifier‑guided RL.
- Integrate research outputs with systems, simulation, and infrastructure partners to deploy real‑world solutions.
- Document findings, publish papers, and present to both technical and non‑technical stakeholders.
Required Skills:
- Strong research foundation in LLMs, reasoning, or autonomous agents.
- Experience with SFT, RLHF/DPO, verifier‑guided RL, or comparable training methodologies.
- Proven ability to design and evaluate long‑horizon behaviors and multi‑step reasoning pipelines.
- Comfortable operating in an interdisciplinary, fast‑paced R&D setting.
- Proficiency in Python, deep‑learning frameworks (e.g., PyTorch, TensorFlow), and research‑grade code repositories.
- Ability to critically assess model safety, robustness, and real‑world applicability.
Required Education & Certifications:
- PhD or M.S. in Computer Science, Machine Learning, Artificial Intelligence, Robotics, or a closely related field.
- Scholarly publications in top-tier venues (ICML, NeurIPS, ICLR, etc.) preferred.
San francisco bay, United states
On site
26-01-2026