Mid-Level Full-Stack Developer (Python + HTLM/JavaScript)

About the Role

We are looking for a Mid-Level Full-Stack Developer to join a fast-moving startup client building an AI-based product in the US healthcare. You’ll own features end-to-end: a FastAPI backend and server-rendered, dynamic HTML on the frontend.

This is a hands-on role for someone who ships. The team keeps processes light and bureaucracy minimal, and delivery expectations are high. AI-assisted development is central to how the team works, so the candidate needs a strong sense of AI coding. That means knowing how to use AI tools to move fast, and knowing when to question, refine, or reject what they produce.

At Vitamin Software, we partner with ambitious companies to build software that solves real-world problems. For this engagement, you’ll be embedded with a startup client working at the forefront of AI products.

What You’ll Do

• Build and maintain backend services and HTTP APIs in Python with FastAPI.

• Build interactive UI in server-rendered HTML: Jinja templates, HTMX, Alpine and Tailwind. The work is dynamic pages, not a heavy single-page app.

• Add LLM features through the existing model layer: prompts, structured output, failure cases, and an eval for each prompt you add or change.

• Use an AI coding tool every day, and review what it writes for correctness, security and fit with the code around it before it is committed.

• Design PostgreSQL tables and queries, including the ones that have to stay fast.

• Move slow or multi-step work onto the background worker, Temporal, instead of holding an HTTP request open.

• Work directly with founders and product and take a ticket from the written spec to something merged.

• Raise concerns early about scope, risk or a better approach, and come with a proposal.

Requirements

• 3–5 years of professional experience in full-stack or backend-leaning web development.

• A strong Python and FastAPI background is enough. The codebase is FastAPI.

• Solid JavaScript, HTML and CSS, including dynamic interfaces without a React-style frontend.

• You have used an LLM inside a product or your own workflow, and you can tell a good model response from a bad one.

• You use an AI coding assistant and you reject generated code that is wrong, insecure or larger than the change needs.

• You have used Linear or another issue tracker.

• You understand HTTP APIs, sessions, and how a multi-tenant web app keeps one customer’s data off another’s.