KenkoTech Futures · 11 hours ago
Principal Research Scientist, RL & Agentic Systems
KenkoTech Futures is a well-funded, research-driven AI company focused on large-scale reasoning models and biological discovery. They are seeking a Principal Research Scientist to lead reinforcement learning efforts for large reasoning models applied to biology and drug discovery, working closely with a senior research team.
Responsibilities
Own and drive reinforcement learning for large reasoning models applied to biology and drug discovery
Design and run large-scale training loops, including reward design, curriculum learning, and evaluation
Improve stability, reliability, and long-horizon reasoning in large models
Build agentic behaviors such as planning, tool use, and multi-step scientific reasoning
Apply these systems to real discovery problems, including target identification and molecule optimization
Work closely with a small, senior research team to define technical direction and research priorities
Operate end-to-end: experimentation, training, evaluation, and iteration at production scale
Qualification
Required
Strong, hands-on experience with reinforcement learning for large models (or extremely adjacent systems)
Proven experience training models at scale - not learning RL-on-LLMs for the first time
Senior-level autonomy and ownership; able to operate independently in a small team
Strong intuition for large-model training dynamics, failure modes, and evaluation
Exposure to biological, chemical, or drug discovery workflows (proteins, antibodies, chemistry, or related domains)
Comfort working in ambiguous, research-heavy environments where the problem is still being defined
Preferred
Experience with autonomous agents or 'AI scientist'–style systems
Protein, antibody, or molecular modeling exposure
Experience designing evaluation frameworks for reasoning or agentic models
Background in applied research labs or frontier AI teams
Company
KenkoTech Futures
健康 (kenko) – The balance of vitality and harmony in body and mind.
Funding
Current Stage
Early StageCompany data provided by crunchbase