You will develop and train protein and antibody foundation models and invent new methods where published approaches fall short. Working inside a modeling workstream, you will implement architectures, train at scale on dedicated LillyPod capacity, and iterate model training to improve performance with feedback from experimental validation. This is a hands-on research role embedded in a multidisciplinary team. You will solve key technical problems within a modeling workstream, contribute novel research ideas, and help translate advances in AI into measurable impact on therapeutic discovery.
· Develop model architectures, implementing, adapting, and where needed inventing approaches for protein design across sequence-structure generation, epitope conditioning, and confidence prediction.
· Contribute novel technical ideas and research directions, helping advance the state of the art in foundation models and generative AI for biotherapeutics.
· Run training at scale, executing training and fine-tuning experiments on LillyPod while tracking configurations, checkpoints, and results for reproducibility.
· Curate training data, preparing internal structure, sequence, and assay data, including processing and interpretation, in partnership with subject matter experts.
· Evaluate and calibrate, building evaluations that connect model outputs to experimental results.
· Diagnose model failures, investigating failure modes, identifying root causes, and developing improvements to model architectures, data strategies, or training approaches.
· Partner cross-functionally, working closely with peers, protein scientists, and automation colleagues to benchmark model performance.
· Communicate results, presenting clearly to both computational and wet-lab audiences and publishing at external conferences and in the scientific community.