Research Scientist

Goliath Partners Inc.; United States; On-site; USD 350,000 - 400,000; Full time

Benefits offered by this job

Equity compensation

Unlimited AI tools budget

Job summary

Goliath Partners Inc. is seeking a reinforcement learning specialist to own the full RL loop and build evaluation and reward layers for frontier AI models using clinical and genomic context.

You will design and run post-training experiments with SFT, RLVR, RLHF/RLAIF, and reward modeling, building scalable pipelines and debugging large ML codebases to move ideas into working prototypes.

Qualifications

• 1+ years of reinforcement learning focused experience.
• Experience designing agent environments and evaluation harnesses.
• Experience with language models and post-training workflows (SFT, RLHF/RLAIF).
• Ability to assess data quality, signals, and clinical relevance.

Responsibilities

• Build comprehensive RL environments including tasks, actions, tools, verifiers, and evaluators.
• Run post-training experiments on LMs and agents using RL techniques.
• Train and evaluate multi-step agents on complex patient histories and data.
• Analyze failures to improve data, rewards and feedback signals.
• Move concepts to prototypes by debugging large ML codebases.

Skills

Reinforcement Learning

Agent Environments

Post-Training LM Evaluation

Reward Modeling

Clinical Data Understanding

Job description

Our client is a heavily backed AI Startup that is revolutionizing healthcare AI to radically improve drug discovery and with the goal to cure cancer, they are building the foundational data layer and pre & post-training simulation environments where the next generation of medical AI models learn. They are seeking someone to own the full reinforcement learning loop and build the crucial evaluation and reward layer for frontier AI models using real-world clinical and genomic context.

Role & Impact

• Build comprehensive RL environments, encompassing task design, action spaces, tool interfaces, verifiers, and evaluation harnesses.
• Run sophisticated post-training experiments on language models and agents utilizing techniques like SFT, RLVR, RLHF/RLAIF, and reward modeling.
• Train and evaluate multi-step agents to navigate complex patient histories and unstructured medical data to capture correctness in clinical reasoning.
• Analyze model failures to continuously improve the data, rewards, and feedback signals that dictate what these systems learn.
• 1+ years of highly technical experience focusing on reinforcement learning, agent environments, LM post-training, or related ML systems.
• Strong judgment around data quality, capable of assessing signal fidelity, coverage, and clinical relevance to support meaningful training tasks.
• The ability to build scalable pipelines, debug complex training runs within large ML codebases, and move rapidly from research concepts to working prototypes.

Compensation: Up to $400,000 base + up to 1.0% equity + comprehensive benefits including unlimited AI tool budgets