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

What you'll do

• Develop post-training strategies, objectives, reward functions, and data-generation approaches for multimodal models.
• Build experimentation, benchmarking, and inference optimization workflows; productionize models and evaluation systems with engineering colleagues.
• Work with drug-discovery scientists to align model development with scientific needs and communicate results.

What they ask for

• PhD in a computational STEM field or equivalent industry experience demonstrating comparable depth.
• Strong Python and PyTorch skills for implementing, training, debugging, and evaluating deep learning models end-to-end.
• Demonstrated experience training large-scale transformer models.
• Experience with supervised or parameter-efficient fine-tuning, reinforcement learning, or systematic hyperparameter optimization; strong engineering practices and comfort with modern ML infrastructure.

Nice to have

• Experience with multimodal or multi-task architectures, training or inference optimization, distributed training, or HPC.
• Familiarity with biomedical, chemical, or biological data domains.

Tools

Python, PyTorch, LoRA, Optuna, Ray Tune, Docker, CUDA, Kubernetes, Weights & Biases

AI focus

Post-training large multimodal foundation models for drug discovery, including fine-tuning, reinforcement learning, reward optimization, evaluation, and inference deployment.