Principal Computational Scientist

About Singleton BioSingleton Bio is an early-stage therapeutics company using generative AI models of human tissue to discover therapeutic targets in immunological and inflammatory diseases. Our mission requires strategic thinking, nimbleness, and deep expertise in several areas including computational biology, immunobiology, in vitro and in vivo pharmacology. This new startup company is recruiting for talented, experienced, and passionate individuals who are looking to meaningfully contribute to the creation of new medicines. Our Cell Interaction Foundation Model (CIFM) or “virtual tissue model” platform is a geometric graph neural network trained on spatial genomics data from millions of human cells. Built on more than a decade of virtual cell and tissue modeling research, the model learns the relationship between a cell’s expression state and its tissue microenvironment, and is used to simulate how tissue responds to perturbations at scale.Singleton is a fast-paced, dynamic environment where teammates will have broad exposure to the intersection of artificial intelligence-driven tissue models and drug development. We are rapidly building the Company and are looking for scientists to join us on this mission. Singleton Bio, located in South San Francisco, CA, is seeking an exceptional individual to join our team as a Principal Computational Scientist.Position SummaryWe are seeking a Principal Computational Scientist to help lead the development of the virtual tissue model as Singleton’s core discovery platform. As the company’s first computational hire, you will take ownership of the model, training data, in silico perturbation screens, and analysis pipeline to support target identification, next-generation model architecture, and downstream drug development applications. You will work closely with all members of the Singleton team including the CTO, who leads model architecture; our Head of Computational Biology; and the pharmacology team.This role suits a scientist who has trained deep learning models on single-cell or spatial data and has also taken responsibility for how those models are evaluated and used.Key ResponsibilitiesOwn the CIFM codebase and trained models on Singleton infrastructure, including training, inference, checkpointing, and versioningRetrain and adapt the model on Singleton’s growing spatial transcriptomics corpus, and assess where additional data or model capacity improves performanceDesign and run the benchmarks used to evaluate model performance, including baseline comparisons, patient- and section-level held-out splits, and out-of-distribution tests across tissues and assay platformsCharacterize model limitations and uncertainty, and communicate them clearly to internal teams and external partnersRun in silico perturbation screens to nominate targets and combinations for our disease programsWork with pharmacology team on the model analysis strategy to yield mechanistic hypotheses and inform cell/tissue-based validation experimental approaches; use experimental results to improve the modelPresent analyses to leadership, partners, and investors clearly and rigorouslyContribute to scientific strategy, publications, and patentsHelp hire and mentor future members of the computational teamQualificationsRequiredPh.D. in computational biology, machine learning, bioinformatics, or a related quantitative field, with 8+ years of relevant industry or applied research experience (exceptional candidates with fewer years will be considered)Demonstrated experience designing and training deep learning models on biological data in PyTorch, including multi-GPU or distributed trainingDeep hands-on experience with single-cell and spatial transcriptomics data, including quality control, normalization, batch correction, and integration across platformsTrack record of taking research code to reproducible, maintainable software used by othersStrong software engineering fundamentals: Python, Git, Linux, containerized environments, and cloud computing (GCP or AWS)Ability to critically evaluate model performance and analytical choices, including rigorous benchmarking against appropriate baselinesDemonstrated ability to independently lead scientific initiatives and collaborate across computational and experimental teamsExcellent written and verbal communication skills, including presenting complex results to non-specialist audiencesPreferredExperience with graph neural networks or geometric deep learningExperience with self-supervised, masked, or generative modeling approachesExperience with perturbation modeling or perturbational genomics (e.g., Perturb-seq, CRISPR screens, in silico perturbation)Experience with imaging-based spatial platforms (e.g., 10x Xenium, NanoString CosMx, Vizgen MERSCOPE) and their technical limitationsExperience with structure-guided protein design (RFdiffusion, ProteinMPNN, AlphaFold/Boltz, RosettaFold, etc.) and familiarity with molecular dynamics simulation tools Knowledge of immunology, tissue microenvironment biology, or ligand–receptor signalingDrug discovery experience, including translating computational predictions into experimental validationPublications in computational biology or machine learning venuesExperience managing cloud compute resources and budgets for large-scale training