Iambic Therapeutics
Machine Learning Scientist — Large multimodal models
About this role
Join Iambic Therapeutics as a Machine Learning Scientist to research and develop Enchant, a large multimodal transformer model for drug discovery. You'll advance model architectures, optimize training and inference at scale, and collaborate across ML and scientific teams to deploy AI technologies that impact therapeutic development.
What you'll do
- Design and implement architectural improvements to multimodal transformer models for biomedical applications
- Optimize training pipelines and inference techniques for efficient deployment across GPU clusters
- Develop benchmarking frameworks to evaluate model quality across modalities and downstream tasks
- Collaborate with ML engineers and computational scientists to deploy models into discovery workflows
- Partner with chemists and biologists to align model development with drug discovery requirements
- Mentor junior team members and contribute to research roadmap strategy (depending on level)
What they're looking for
- Python and PyTorch
- Transformer model training at scale
- Deep learning implementation end-to-end
- Training and inference optimization
- ML infrastructure (Docker, CUDA, Kubernetes, Weights & Biases)
- Multimodal and multi-task architectures
- Distributed training
- Reproducible experimentation and clean code practices
Opens the application — the Jobs AI extension fills it for you. Set up autofill
Opens the official application on the employer’s site. No login required.
Iambic Therapeutics
Iambic Therapeutics builds Enchant, a multimodal transformer model designed to accelerate drug discovery through AI-powered clinical prediction and therapeutic decision-making. The company is hiring Machine Learning Scientists to develop and deploy the model, including roles focused on fine-tuning for clinical tasks, building biomedical data pipelines, and advancing model architectures and training at scale.
View all jobs at Iambic TherapeuticsLikely interview questions
- Walk us through a transformer model you've trained at scale. What were the key bottlenecks you encountered during training, and how did you optimize for efficiency?
- Describe your experience with multimodal architectures. How have you approached combining different data types (e.g., images, text, graphs), and what design decisions did you make?