Active Learning & Multi-Objective Optimization: Design and establish Active Learning pipelines for multi-objective optimization balancing affinity, specificity, stability, immunogenicity, and manufacturability; include multi-property guidance, Pareto-optimal search strategies, and uncertainty quantification.
Reward & Surrogate Modeling: Design and train reward models and discriminative classifiers (e.g., affinity ranking, stability prediction, developability scoring) as objective functions for optimization loops.
Reinforcement Learning for Generative Model Alignment: Develop and implement RL strategies (PPO, DPO, reward-weighted approaches) to fine-tune generative models (autoregressive transformers, diffusion models) toward biologic sequences with desired therapeutic properties; assess when RL vs Bayesian Optimization/Active Learning is warranted.