About the job
Altos Labs is building a world-class AI ecosystem to solve the most complex problems in human biology. You will directly design and build high-performance, scalable solutions that unify high-dimensional biomedical imaging with molecular and language data. By implementing large-scale multimodal data fusion, you will move beyond simple image analysis to create predictive models that map across biological domains. You will be hands-on with the data and the code, collaborating with our engineering team to ensure these models are scalable, efficiently trainable on distributed cloud infrastructure, and accessible to our global research community.
Responsibilities
Model Development: Design, code, and train large-scale foundation models (e.g., Vision Transformers, Multimodal LLMs) that can embed spatial data and integrate multiple modalities.
Hands-on Data Fusion: Implement innovative cross-domain mapping and fusion strategies to synchronize heterogeneous biological datasets.
Scaling & Training: Build and manage high-performance ML pipelines capable of processing petabyte-scale image repositories and multi-omics streams in a cloud environment.
Technical Collaboration: Work directly in the trenches with experimental scientists and software engineers to translate biological complexity into performant code and reliable distributed systems.
Qualifications
Minimum
Education: PhD in Computer Science, AI/ML, Biomedical Engineering, or a related quantitative field.
Hands-on CV Expertise: Deep experience building and deploying modern Computer Vision architectures (Vision Transformers, U-Nets, Self-Supervised Learning).
Distributed Training: Proven experience training and fine-tuning large models at scale using frameworks like PyTorch Distributed, DeepSpeed, or Jax.
Programming Mastery: Expert-level Python skills, with a focus on building production-ready machine learning code and large-scale data management systems.
Scientific Contributions: A track record of technical contributions via high-impact publications (CVPR, ICCV, NeurIPS, etc.) or significant contributions to open-source ML frameworks.
Preferred
Direct experience with Multimodal Fusion (e.g., aligning image embeddings with transcriptomic or proteomic data).
Proficiency with cloud-native AI tools (AWS/GCP, Kubernetes, Docker) and building automated MLOps workflows.
Experience handling the unique noise and sparsity of biological data.