Research Scientist

The role

A Research Scientist will set and pursue a research agenda for reliable long-horizon agents in enterprise environments. The role involves inventing methods for post-training agents, designing memory architectures, and developing evaluation methodologies.

What you will do

• Set and pursue a research agenda on reliable long-horizon agentic behavior
• Invent and validate methods for post-training agents on transformation work
• Define memory architecture for long runs
• Develop evaluation methodology to predict customer-observed correctness
• Study how multi-agent systems fail
• Work on the verification problem directly
• Solve how a system represents an enterprise to itself
• Investigate post-training compute and data quantity returns
• Turn findings into things that ship
• Publish papers, technical reports, and open-source artifacts

Required qualifications

• MS or PhD in CS, ML, statistics, math, physics, or a related field, or research experience of comparable depth
• Track record of original research in ML: publications at NeurIPS/ICML/ICLR/ACL-tier venues, influential open-source work, or results inside a lab that moved a frontier system
• Deep expertise in at least one of: post-training and RL for LLMs, agent memory and long-context reasoning, knowledge representation and structured reasoning, or evaluation methodology
• Hands-on ability to write code, run experiments, and read logs
• Exceptional research judgment
• Ability to hold rigor and urgency at once
• Ability to write clearly and explain technical arguments

Nice to have

• Experience owning a research direction end to end at a frontier lab or a strong academic group
• Published work on agents, tool use, RL for LLMs, code generation or repair, reasoning, or evaluation
• Experience with verifiable-reward RL, or with the failure modes of reward models where verification is incomplete
• Experience with knowledge representation, ontologies, or neurosymbolic approaches to structured domains
• Experience with training runs at meaningful scale
• A history of mentoring researchers and engineers

About the company

Tessera Labs is an AI platform designed as a transformation engine for large enterprises. It is a multi-agent platform that understands enterprise process, data, and code as a connected system.