Staff Software Engineer (AI)

About the role

We're seeking an experienced and highly capable AI engineer who is motivated to meaningfully improve the way healthcare is administered in the United States. You'll take ownership of developing a significant part of the Onos platform, including the development of models for assessing the level of appropriate care for patients. As an early team member, you'll be expected to wear multiple hats, including acting as a backend/data engineer, while ensuring excellent outcomes for our customers. This role is a hybrid role based in San Francisco, where you'll be expected to work at our office in person 2-3 times a week. What you'll be doing at Onos: Develop LLM/NLU systems to process and extract meaningful information from clinical notes and medical documents, classify patients according to level-of-care guidelines, and make accurate recommendations Establish best practices for LLM/AI systems, benchmarking/evaluation frameworks, and model governance to improve reliability, maintainability, and scalability Build data pipelines that scale efficiently while maintaining strict data privacy and security standards Collaborate with backend engineers to integrate AI/ML capabilities seamlessly into the Onos platform Technical Challenges At Onos: Build and operationalize AI/data pipelines to analyze medical records to streamline clinical assessments and healthcare quality reviews Benchmark LLM systems to more accurately and reliably extract evidence from medical records and classify patients' level-of-care recommendations Develop and optimize a system that ingests complex medical standards of care documents and evaluates provider adherence to guidelines Design explainable AI solutions that provide transparency into model decisions for healthcare professionals Tech Stack: Infrastructure/Systems: AWS (ECS, Bedrock, Cognito, etc.), Docker, Github Actions Languages/Frameworks: Python, Django, Celery, django-ninja, django-tenants Database/Storage: PostgreSQL (AWS RDS), S3 Development Tools: Github, Jira, CoderabbitAI, Tusk, Claude