Machine Learning Scientist — Large Multimodal Models (Post-Training)

About this role

Join Iambic's Enchant team to develop post-training methods for large multimodal foundation models applied to drug discovery. You'll design and evaluate cutting-edge approaches like reinforcement learning and fine-tuning strategies, build scalable training infrastructure, and collaborate across ML, engineering, and chemistry teams to advance therapeutic innovation.

What you'll do

• Research and develop post-training strategies for multimodal foundation models, including reinforcement learning, fine-tuning, and emerging optimization methods
• Design reward functions, training objectives, and evaluation protocols for post-training approaches on large-scale LLMs
• Build experimentation workflows and hyperparameter optimization pipelines for efficient exploration of training recipes and configurations
• Develop inference optimization techniques and deployment strategies for evaluation and interactive discovery workflows
• Create and maintain rigorous benchmarking frameworks measuring model quality across modalities, tasks, and drug discovery use cases
• Collaborate with ML engineers and domain scientists to productionize models and ensure alignment with therapeutic needs

What they're looking for

• Python and PyTorch
• Large-scale transformer model training
• Reinforcement learning (RLHF, RLAIF, PPO, GRPO)
• Parameter-efficient fine-tuning (LoRA) and supervised fine-tuning
• Hyperparameter optimization (Optuna, Ray Tune)
• ML infrastructure (Docker, CUDA, Kubernetes, Weights & Biases)
• Training and inference optimization (quantization, mixed precision, distributed strategies)
• Strong software engineering practices and reproducible experimentation