Backbone Systems · San Francisco
OtherEngineeringFullTime
Care should be paid for the moment it’s delivered. At Backbone, we’re making that possible for the first time by building the clinical AI layer for healthcare payments.
Today, U.S. healthcare wastes more than $350B yearly on administrative work that has nothing to do with patient care: prior authorizations, denials, appeals, and the staff hired on both sides of the aisle to adjudicate and determine payment for services rendered. As a result, patient care is delayed, clinicians burn out, margins crater, and hundreds of billions of dollars that should be funding care end up funding the friction itself.
Backbone is dismantling this friction, building intelligent financial rails atop clinical AI research to facilitate instant payments between payers and providers. We're backed by Andreessen Horowitz (a16z), Lightspeed, Hanabi, and a roster of operators and clinicians who have built and run the largest payers, providers, and healthcare technology platforms in the country. Our founder/CEO, Manan Shah, holds a BS/MS in Math & CS from Stanford and led ML infrastructure at Sequoia-backed Kumo AI, where he built graph neural networks serving DoorDash, Reddit, and Coinbase.
Backbone’s revenue is growing quickly, customer demand is outpacing our team, and the road ahead is bright.
We’re solving applied research problems in clinical intelligence. Backbone’s core systems, based on post-trained open-weights models and systems of closed-weights models, read charts, claims, denials, appeals, clinical criteria, payer policies, and quality metrics, then make accurate, defensible decisions in production. We also leverage continually improving browser-use and computer-use models to navigate payer portals, EHRs, practice management systems, and the messy software layer that healthcare runs on.
This is not a pure research role. Alongside core model improvements, you’ll take ambiguous research problems, turn them into evals, improve model behavior, and work with engineers to put those improvements into production.
You will:
Build evals for clinical reasoning, chart understanding, claims review, authorization logic, denial analysis, appeal generation, and payer policy interpretation
Improve Backbone’s ability to read and reason over real clinical and financial data, with work across model behavior, prompting, fine-tuning, retrieval, structured reasoning, verification, and evaluation
Build and improve agentic browser-use and computer-use capabilities for payer portals and healthcare software
Translate product feedback from live provider and payer workflows into research problems, partnering with backend engineers to productionize research improvements
New grad through ~5-6 years of experience, with experience with ML research, language models, NLP, evals, reinforcement learning, agentic systems, or model improvement
Clear evidence of technical depth through papers, projects, internships, open-source work, competitions, or production ML systems. Infrastructure or systems experience is a major plus.
Excitement about building applied research systems where the output touches real workflows, real dollars, and real patient impact.
Healthcare experience is helpful, but not required. Curiosity, rigor, and taste matter more.
This is a chance to join early and help define both the product and the engineering culture.
You’ll work on hard, practical problems: turning fragmented healthcare workflows into software that actually helps people do their jobs faster and better. Your work will touch product decisions, system architecture, user experience, and the core automation engine behind Backbone.
We offer competitive compensation, meaningful ownership, and the opportunity to build foundational technology in a massive, broken, high-impact industry.