This list contains only the countries for which job offers have been published in the selected language (e.g., in the French version, only job offers written in French are displayed, and in the English version, only those in English).
As part of our team, you will help to accelerate and optimize our progress in developing unified, multi-modal generative foundation models for multiscale biology. You will be an integral part of our multidisciplinary teams building the computational platforms that will enable Altos to achieve its mission. In this role, you will partner and collaborate with other multidisciplinary Scientists and Engineers across the Institute of Computation to design, build, and scale state-of-the-art foundation models that tackle biological questions and aid in the discovery of novel interventions for aging and disease. You will focus on the synthesis of unstructured multimodal signals with the structured relational data and knowledge graphs that represent biological reality.
Job Responsibility:
Pre-train and fine-tune large-scale machine learning systems using multimodal biological data, natural language, and structured relational inputs
Architect and implement novel hybrid models that integrate Large Language Models (LLMs) with Graph Neural Networks (GNNs) for multi-hop reasoning over biological knowledge graphs
Develop Relational Foundation Models (RFMs) that enable zero-shot predictive tasks over heterogeneous, multi-table biological datasets
Lead the design of efficient data loading strategies and distributed training recipes (e.g., FSDP, DeepSpeed) to train models across multiple GPU nodes
Gain insights into model performance based on theory, deep research, and the mathematical underpinnings of set-invariant and graph-structured architectures
Apply strong coding experience to model development and deployment, ensuring research prototypes transition into reliable, scalable production systems
Stay up-to-date on the latest developments in deep learning—including native early-fusion and Mixture-of-Experts (MoE) architectures—and apply this knowledge to Altos' research
Mentor junior staff while maintaining a high individual technical contribution to the core research ecosystem and peer-reviewed publications
Requirements:
PhD in Computer Science, Machine Learning, or a similar quantitative field with 5+ years of relevant work experience in academic or industry settings
Prior experience in developing and implementing novel generative AI models, specifically in multimodal integration, GraphRAG, or relational deep learning
Deep understanding of Machine Learning principles and how they apply to diverse architectures like Transformers, GNNs, and diffusion models
Very strong programming skills in Python and deep learning libraries (e.g., PyTorch, JAX, Hugging Face Transformers/Accelerate)
Proven experience with multi-GPU and distributed training at scale (e.g., DDP, FSDP, DeepSpeed, Megatron, or Ray)
Strong track record of published, peer-reviewed innovative AI/ML research at top-tier conferences (NeurIPS, ICML, ICLR, CVPR)
Nice to have:
Familiarity with tabular foundation models (e.g., TabPFN) and in-context learning strategies for structured data
Specific experience in native multimodal modeling (early-fusion) or the synthesis of LLMs and Knowledge Graphs
Track record of ML applied to biological data, such as NGS data (RNA-seq, ATAC-seq), biological imaging (microscopy, IF), or spatial transcriptomics
Experience in optimizing large-scale inference via quantization, distillation, or memory-efficient attention mechanisms