Senior AI Scientist

Remote, United StatesFull-time$180,000 – $260,000 / Annual

Atria is a revolutionary preventive healthcare movement built around one question: what is biologically possible, now and in the future
Atria delivers comprehensive, personalized preventive care designed to extend healthy lifespan through meticulous screening, advanced therapeutics, and extraordinary compassionate care
Atria is being built ground-up as a technology-forward, AI-native preventive medicine platform

Own the medical modeling roadmap that powers Atria's clinical AI agenda, from domain-specific models to foundation models for preventive medicine
Work on a data estate few organizations have access to: whole-genome sequencing, whole-body MRI, longitudinal labs, wearables, and multi-generational family records
Take the best of the open-source ecosystem and turn it into models that materially outperform off-the-shelf options on Atria's clinical problems

Own the medical modeling roadmap, identifying and building the model families and training approaches that move the clinical needle
Survey the open-source and clinical/biomedical foundation model landscape and make defensible fine-tune, distill, or build decisions
Lead fine-tuning and post-training work: SFT, LoRA/QLoRA, DPO and preference-tuning, RLHF/RLAIF, continued pre-training, and distillation
Curate high-quality, PHI-safe training datasets from Atria's clinical data with rigorous sampling, labeling, and contamination checks
Design and run rigorous evaluation against held-out clinical benchmarks and frontier baselines, with calibration and subgroup analysis

4+ years hands-on experience training and fine-tuning modern deep learning models, with shipped or published work
Graduate degree (PhD preferred) in computer science, ML, computational biology, or biomedical informatics, or an equivalent open-source track record
Deep, current fluency with SFT, PEFT (LoRA, QLoRA), preference tuning, distillation, and continued pre-training
Strong Python and PyTorch with hands-on experience in the Hugging Face ecosystem (transformers, datasets, PEFT, TRL, accelerate)
Healthcare or clinical/biomedical foundation model experience a strong plus