Senior Software Engineer, Agentic AI

Heartflow•San Francisco, CA

23h

About The Position

Heartflow is a medical technology company advancing the diagnosis and management of coronary artery disease, the #1 cause of death worldwide, using cutting-edge technology. The flagship product—an AI-driven, non-invasive cardiac test supported by the ACC/AHA Chest Pain Guidelines called the Heartflow FFRCT Analysis—provides a color-coded, 3D model of a patient’s coronary arteries indicating the impact blockages have on blood flow to the heart. Heartflow is the first AI-driven non-invasive integrated heart care solution across the CCTA pathway that helps clinicians identify stenoses in the coronary arteries (RoadMap™Analysis), assess coronary blood flow (FFRCT Analysis), and characterize and quantify coronary atherosclerosis (Plaque Analysis). Our pipeline of products is growing and so is our team; join us in helping to revolutionize precision heartcare. Heartflow is a publicly traded company (HTFL) that has received international recognition for exceptional strides in healthcare innovation, is supported by medical societies around the world, cleared for use in the US, UK, Europe, Japan and Canada, and has been used for more than 500,000 patients worldwide. Coronary artery disease is the world's leading cause of death - and we're changing that. Heartflow builds AI that gives clinicians a personalized, precision 3D model of a patient's heart from a single CT scan, helping them diagnose and treat CAD earlier, more accurately, and without invasive procedures. To date, we've helped clinicians manage care for nearly 650,000 patients across more than 1,800 institutions worldwide. Every line of code we ship has a real person on the other side of it. We're looking for a Senior Software Engineer to help lead our agentic AI platform from the inside out. This is a small-team, high-ownership role - you'll be one of a handful of engineers working at the intersection of cutting-edge AI and FDA-regulated medical device software, with direct influence on how Heartflow builds and ships software internally. This isn't a feature team role. You'll be laying the infrastructure that the rest of engineering builds on top of, working shoulder-to-shoulder with world-class deep learning researchers, clinical scientists, and a leadership team that's deeply invested in getting this right. If you want your work to matter beyond the sprint board, this is that job.

Requirements

Strong proficiency in Python and TypeScript for building end-to-end agentic applications.
Deep experience with FastAPI for backend development and ML model integration.
Proficiency in React (or similar modern frontend frameworks) for building agentic UIs and MCP client.
Hands-on experience with AWS and Infrastructure as Code (Terraform).
Experience with Agile development methodologies and project management tools.
Deep understanding of LLM API integration (Anthropic Claude, OpenAI), prompt engineering, and RAG orchestration.
Production experience building deterministic tool integrations for non-deterministic LLMs using MCP server/client architecture.
A track record of moving beyond manual testing to automated CI/CD-integrated evaluation using platforms like Langfuse, Braintrust, or Promptfoo.
Experience with Docker, RESTful APIs, Git, and CI/CD pipelines for deploying agents at scale in secure, isolated environments.
Ability to communicate complex AI concepts clearly - in writing and in conversation - to audiences ranging from engineers to regulatory reviewers.
Familiarity with data privacy requirements in healthcare settings (HIPAA, PHI handling).
Experience driving technical vision and leading architectural decisions; ability to mentor and lead other engineers.
Bachelor's degree in Computer Science, Engineering, or related discipline, or equivalent experience.
7+ years of relevant industry experience.

Nice To Haves

Experience with high-performance languages (Go, Rust, C++) for building concurrent, low-latency API infrastructure.
Advanced experience fine-tuning open-weight models (LoRA, PEFT) or deploying local models with PyTorch or Hugging Face.
Hands-on experience with advanced orchestration tools like LangGraph, Pydantic AI, or multi-agent frameworks; browser automation with Playwright.
Experience with AI-specific adversarial testing and prompt security - red teaming, STRIDE threat modeling, or Promptfoo evaluations.
Familiarity with serverless architecture (Lambda), Kubernetes, and vector databases (Pinecone, pgvector) for scalable agent deployments.
Background in MedTech regulatory frameworks - FDA 21 CFR Part 11, IEC 62304, GxP, ALCOA+, or SBOM/VEX generation.

Responsibilities

Own the core AI platform infrastructure. You'll build and maintain the shared libraries, MCP tool integrations, and deployment pipelines that power our internal agentic systems - the foundation other product teams depend on.
Apply AI to how we build regulated software. One of the most distinctive and genuinely novel parts of this role: you'll use LLMs to accelerate Heartflow's own software development lifecycle - parsing product requirements, mapping them to software modules, and auto-generating test cases - so our engineering team can move faster without sacrificing the compliance rigor that FDA-cleared medical device software demands.
Design Human-in-the-Loop workflows that actually work. Agentic AI in a regulated environment means humans stay in the loop in meaningful ways. You'll architect HITL workflows that keep AI-generated outputs aligned with FDA requirements - not as an afterthought, but as a core design principle.
Build and evolve our knowledge retrieval architecture. You'll design the pipelines that let our multi-agent systems find, reason over, and synthesize complex clinical and regulatory data - evolving from RAG foundations toward graph-based and agentic retrieval approaches as the needs grow.
Connect Claude to the tools our teams actually use. You'll develop and maintain MCP host environments, client applications, and servers that securely bridge our AI systems with enterprise platforms like GitHub, Jira, Salesforce, and Slack.
Build evaluation systems that prevent regressions before they reach clinicians. You'll create automated evaluation harnesses - LLM-as-a-judge pipelines integrated into CI/CD - so we continuously validate AI output quality and catch problems early. No manual vibe-checks.
Build sophisticated agentic frameworks. You'll combine foundation models with operational patterns like ReAct and Plan-and-Solve architectures, managing agent memory via vector databases to support complex, multi-step workflows.
Create internal developer tooling that unlocks the broader team. You'll build the APIs, sandboxes, and DX infrastructure that let product engineers across Heartflow safely integrate AI into their workflows - without needing to become AI specialists themselves.
Mentor and translate. You'll help teammates grow in agentic AI practices and serve as a bridge between engineering and clinical, regulatory, and product stakeholders - helping non-technical audiences understand what AI can and can't do in a regulated environment.