Senior Software Engineer

Skylight Health is transforming specialty care through a value-based care platform that connects patients, clinicians, and payors. The Senior Software Engineer will own core development of a referral platform, building agentic workflow orchestration, backend and frontend infrastructure, healthcare data integrations, and scalable systems in collaboration with technical, product, data science, and clinical teams.

Responsibilities:

• Agentic referral workflow development : Build and maintain healthcare referral workflow definitions, agent coordination logic, and integrations using browser agents within the platform. Continue to buildout our outbound agent calls and text conversations with patients and clinicians. Incorporate internally developed intelligent routing and multispecialty intelligence into the referral engine
• Web app infrastructure : Develop backend APIs (Node.js, REST/GraphQL), web UI components (React or similar), and data models that create seamless human-on-the-loop and in-the-loop experiences for users
• Healthcare claims data integration & analytics : Use healthcare claims data from CMS and other sources to inform workflow behavior, build new data pipelines into the app, and maintain existing analyses that power workflow decisions
• Technical debugging & optimization : Troubleshoot workflow failures, optimize data retrieval from healthcare sources, and improve platform reliability and performance at scale
• Cross-functional collaboration : Work with data science, engineering, product, and clinical teams to understand healthcare referral workflows and translate requirements into technical solutions
• Onboard on the platform codebase, understand the current referral workflow architecture, and get familiar with CMS healthcare claims data integration patterns
• Build relationships with the CTO, engineering team, and product stakeholders; understand the current customer use cases and roadmap
• Ship 3 bug fixes or feature refinements to the workflow engine
• Own and ship a meaningful feature or infrastructure improvement (ex. new workflow step type, improved state management, new API endpoint) that directly benefits customers
• Contribute to CMS healthcare claims data pipeline improvements or new analyses that feed workflow decisions
• Establish clear code ownership patterns and begin mentoring on technical decisions across your area
• Deliver a measurable product or infrastructure outcome (ex. new workflow capability, performance improvement, new customer workflow type) that improves customer adoption or referral success rates
• Own a cross-functional initiative (ex. new referral workflow type, integration improvement) end-to-end with minimal oversight
• Contribute to product roadmap planning and help shape the platform’s technical direction

Requirements:

• 5 years of software development in production environments, with demonstrated ability to own systems end-to-end and strong system design judgment
• You have used agents in a production code base and are familiar with SDD for maintaining and extending the code base
• Hands-on experience with agentic or AI systems — you've built, deployed, or maintained agents (LLM-based, workflow orchestration, or similar)
• You understand agent state management, failure modes, and multi-step coordination
• JavaScript/TypeScript mastery — with an agent you ship clean, maintainable TypeScript/JavaScript across backend (Node.js, APIs), frontend (React or similar), and multiple contexts
• You're comfortable with modern web frameworks and tooling
• SQL & data engineering competency — you can understand complex queries, design data schemas, and build/iterate on ETL and data processing pipelines
• Experience in a HIPAA compliant environment
• Experience with large healthcare datasets
• GCP (ex. Cloud Build, Cloud Run, Artifact Registry) or Databricks experience
• PySpark for distributed data processing
• Exposure to healthcare referral, prior authorization, or care coordination workflows
• dbt familiarity for data transformation
• Comfort with PySpark for distributed workloads is valued