Backend Software Engineer - New Bets

About the company

Heidi Health builds AI-powered medical scribe and care-partner software that removes administrative burdens from clinicians.

About this role

WHO IS HEIDI? Heidi is building an AI Care Partner that supports clinicians every step of the way, from documentation to delivery of care. We exist to double healthcare’s capacity while keeping care deeply human. In the last 18 months, Heidi has returned more than 18 million hours to clinicians and supported over 73 million patient visits. Today, more than two million patient visits each week are powered by Heidi across 116 countries and over 110 languages. Founded by clinicians, Heidi brings together clinicians, engineers, designers, scientists, creatives, and mathematicians, working with a shared purpose: to strengthen the human connection at the heart of healthcare. Backed by nearly $100 million in total funding, Heidi is expanding across the USA, UK, Canada, and Europe, partnering with major health systems including the NHS, Beth Israel Lahey Health, MaineGeneral, and Monash Health, among others. We move quickly where it matters and stay grounded in what’s proven, shaping healthcare’s next era. Ready for the challenge? THE ROLE The New Bets squad builds Heidi's next products. Scribe already powers millions of patient visits a week, and Evidence and Remote are scaling behind it. Your job, alongside the New Bets PM and Team, is to build what comes next. This is zero-to-one engineering with a distribution advantage most early products never get. Your v1s don't launch into silence, instead they go in front of clinicians who already open Heidi every day, and who will tell you within a week whether you have built something they need. The squad is small on purpose. You'll build the whole thing: the React front end, the Python services behind it, and the prompts, retrieval and eval harnesses that make an LLM feature safe enough to put in front of a clinician. When a bet needs an onboarding flow, a data migration or a quick internal dashboard, that's you too. There's no one else to hand it to, so you will own the build from PoC through to production. Some of what you build will be killed. Ultimately, most bets don't survive contact with a real clinic, and the winners graduate into staffed product lines that inherit your code. For this to work, the code that dies has to have been fast to write, and the code that lives has to be worth inheriting. To fit in well with this team, you need to be the engineer who ships, ideally with: a few years of production software behind you, real judgment about what's good enough for a prototype versus what has to be right, and the appetite to own more surface area than any normal role would give you. If you're mid-level and impatient about it, this role is a fast-track to exposing yourself to complex problems and pushing towards a Senior level. WHAT YOU'LL DO: - Build new products end to end: React/TypeScript front ends, Python services, and the LLM and voice pipelines in between - Turn a one-page thesis into a working v0 in days, then decide with the PM what deserves a v1 - Make AI features trustworthy and compliant: prompting, retrieval and evals that hold up in a real clinic, not just in a demo - Ship weekly, demo what you shipped, and let real usage settle arguments - Spend time with clinicians to deeply understand how they actually work, and the blockers they face daily - Work with our commercial teams to understand the business case behind what we build - Instrument everything: usage, latency, failure modes. The scale-or-kill call runs on data you built the pipes for - Hand winners over well: when a bet graduates, either you will continue to develop and maintain it, or a staffed team will pick up your codebase WHAT YOU WILL NEED: - 3+ years shipping production software, we care much more about what you've shipped rather than time in seat - Full-stack fluency: Python on the back end, React/TypeScript on the front, comfortable owning everything from the schema to the UI - Working fluency with LLM systems: prompting, retrieval, embeddings, evaluation and the judgment to build around their probabilistic nature - You build with AI tools (Cursor, Claude Code, whatever ships faster) and can show what you've made with them - Cloud fluency (AWS or GCP): you can stand up your own infrastructure and CI/CD and get code to users without a platform team - A testing-in-production mindset: observability and feedback loops, because "works on my machine" doesn't survive a clinic - Product instincts: you ask what the user is trying to do, and you'll argue with the PM when the ticket is wrong - Speed as a habit, with weekly demos and a bias to ship over a bias to plan - Strong opinions, weakly held: you'll shift the room when you're right - Healthcare experience isn't required, but a genuine passion for wanting healthcare to work better is important WHAT DO WE BELIEVE IN? Heidi builds for the future of healthcare, not just the next quarter, and our goals are ambitious because the world’s health demands it. We believe in progress built through precision, pace, and ownership. - Live Forever - Every release moves care forward: measured, safe, and built to last. Data guides us, but patients define the truth that matters. - Practice Ownership - Decisions follow logic and proof, not hierarchy. Exceptional care demands exceptional standards in our work, our thinking, and our character. - Small Cuts Heal Faster - Stability earns trust, speed delivers impact. Progress is about learning fast without breaking what people depend on. - Make others better - Feedback is direct, kindness is constant, and excellence lifts everyone. Our success is measured by collective growth, not individual output. Our mission is clear: expand the world’s capacity to care, and do it without losing the humanity that makes care worth delivering. Why you should Join Heidi 🚀 - Real product momentum. We're not trying to generate interest, we're channeling it. - Equity from day one. When Heidi wins, you win. You'll share directly in