About this role
Isomorphic Labs builds predictive and generative AI models that accelerate scientific discovery to help design novel molecules and develop innovative medicines for disease treatment. The model performance and scaling team focuses on scaling foundational models that transform the biopharmaceutical industry through digital biology. Engineers in this group collaborate closely with scientific colleagues to tackle complex computational challenges at the intersection of machine learning and drug design.
Daily work centers on implementing and optimizing post-training methods for large language models at scale across distributed systems. Team members diagnose communication bottlenecks, evaluate fine-tuning frameworks, and deploy low-precision techniques that balance accuracy with computational efficiency. Finding solutions to these performance puzzles requires a solid grasp of hardware capabilities alongside pragmatic software engineering.
Candidates who thrive here typically bring deep expertise in distributed machine learning infrastructure and a genuine curiosity about biological applications. The environment encourages iterative problem-solving, continuous learning, and cross-disciplinary teamwork to advance rational drug design. Anyone eager to apply advanced engineering skills toward meaningful health outcomes will find a supportive space for their talents.
What you will do
- Implement and scale post-training methods for large language models.
- Partner with research teams to transition novel techniques into robust, production-ready systems.
- Identify and resolve performance bottlenecks and communication overhead within distributed training and inference setups.
- Evaluate and deploy frameworks designed for supervised fine-tuning, reinforcement learning, and model evaluation.
- Deploy low-precision methods to balance execution performance with model accuracy for active drug design programs.
- Prioritize and execute on various performance optimization opportunities across the infrastructure.
What you need
- Demonstrated background working with large-scale distributed training of large language models.
- Proficiency using deep learning machine learning frameworks such as JAX or PyTorch.
- Familiarity with parallelism strategies and collective communication libraries including NCCL.
- Solid understanding of underlying GPU architectures and hardware performance concepts.
- Strong collaborative abilities to work effectively within interdisciplinary teams.
Nice to have
- Experience with general large language model serving stacks.
- Knowledge of accelerator domain-specific languages and compilers like XLA, Triton, Pallas, or CUDA.
- Experience with optimizing machine learning accuracy through low-precision formats.
- Prior background building, deploying, and maintaining production systems on Google Cloud Platform.
- Personal or professional interest in chemistry and biology.
Skills and tech stack
Questions candidates ask
What is the remote work policy?
The company follows a hybrid model requiring employees to come into the office three days a week. These days currently include Tuesday, Wednesday, and a third day depending on the specific team.
What tools and frameworks are used in this role?
The team works with deep learning frameworks like JAX and PyTorch, parallelism strategies, collective communication libraries such as NCCL, and accelerator technologies including XLA, Triton, and CUDA.
What will I be working on day-to-day?
You will implement and optimize large language model post-training methods at scale, diagnose performance bottlenecks in distributed training, and deploy low-precision techniques to support drug design programs.
The application is on the official Isomorphic Labs careers page.
Apply on company site