Advisor - Agent Research

About the Role

We are rebuilding the Design-Make-Test-Analyze (DMTA) cycle, infusing scientific automation with foundation models, multi-agent systems, and robotics to make scientific discovery intelligent, autonomous, and fast. We're seeking a scientist-engineer hybrid to design the learning layer of our scientific agent platform. You will design the environments, rewards, and domain-specific models that enable agents to improve based on experimental feedback. You'll translate wet-lab and computational endpoints into a trainable signal to build models that plan and act against them.

Responsibilities

• Partner with scientists to build autonomous agents that undertake molecule discovery tasks.
• Design and build reinforcement learning (RL) environments that wrap real discovery tasks with appropriate state, action, and termination semantics.
• Curate and engineer reward functions from noisy scientific signal.
• Post-train domain models (SFT, DPO/GRPO/PPO, reward modeling, distillation) on chemistry and biology tasks.
• Integrate learned policies with domain tools (RDKit, molecular graph ML, ELN/LIMS APIs, instrument drivers) so trained models execute real DMTA tasks.
• Build the eval infrastructure: task suites, scoring harnesses, regression tracking, and experiment tracking (e.g., MLflow).
• Represent Frontier AI in the broader AI@Lilly and external AI research community: publish, give talks, review papers, and scout emerging trends.
• Evaluate external vendors, open-source projects, and academic collaborations for strategic fit.

What Success Looks Like

• Trained models that measurably outperform prompted frontier baseline models on internal discovery tasks.
• Reward and evaluation infrastructure that other teams adopt as the default way to measure agent performance.
• Measurable reduction in DMTA turnaround through autonomous planning and execution.
• Seamless transition from prototype to production-deployed AI systems.

Requirements

• PhD (or MS + 3 yrs / BS + 5 yrs equivalent experience) in Machine Learning, Bioinformatics, Cheminformatics, Computer Science, or related discipline with demonstrated wet-lab collaboration or hands-on experience.
• Approximately 1-2 years of demonstrated experience in applying AI/ML in scientific disciplines such as biology, chemistry, neuroscience, or a related field (industry postdoc counts).
• Hands-on experience training or post-training AI models.

Skills

• Proficiency in Python and deep experience with ML/Deep Learning frameworks (e.g., PyTorch, Tensorflow, JAX, HuggingFace).
• Experience with RL and post-training methods (PPO, GRPO, DPO, reward modeling, RLHF/RLAIF) and libraries such as TRL, verl, or equivalent in-house stacks.
• Familiarity with molecular representation learning, generative chemistry, or protein/nucleic acid models.
• Hands-on experience building agentic AI systems (e.g., OpenAI/Anthropic Agent SDK, Langchain, Smol agents).
• Experience designing and shipping end-to-end systems in cloud environments (backend APIs, lightweight frontends, and agentic platforms) - GitHub portfolio a plus.
• Working knowledge of cloud-native (AWS/Azure) pipeline architectures, including Nextflow, Argo on Kubernetes.
• Demonstrable research experience, evidenced by contributions to projects, and ideally through publications in relevant ML/NLP venues (e.g., NeurIPS, ICML, ICLR, ACL, EMNLP).
• Experience mentoring and guiding junior researchers or engineers.

Benefits

• Eligibility to participate in a company-sponsored 401(k) and pension.
• Vacation benefits.
• Eligibility for medical, dental, vision, and prescription drug benefits.
• Flexible benefits (e.g., healthcare and/or dependent day care flexible spending accounts).
• Life insurance and death benefits.
• Certain time off and leave of absence benefits.
• Well-being benefits (e.g., employee assistance program, fitness benefits, and employee clubs and activities).
• Company bonus (depending, in part, on company and individual performance).

Pay

The anticipated wage for this position is $151,500 - $222,200.