PythonReactFlaskFastAPIPostgreSQLAWSTerraformLLM
HEALTH CAN’T WAIT.
Not for symptoms to get worse. Not for a six‑month appointment. Not for a system to catch up. But that’s exactly how healthcare works today. You wait, until you can’t.
Alan exists to end the wait.
Health is a universal right, and we believe this right can only become real when it’s coupled with prevention. We need to stop treating health as something we repair and start treating it as something we build, every day. It’s not solely a question of willpower. It’s the healthcare system itself that needs to work for everyone, in a sustainable way.
So we are building the new standard in prevention insurance. Alan is the first company that integrates insurance, prevention, and care into a single, acclaimed user experience.
We are on an incredible journey to build a global leading company, with a unique culture https://alan.com/en/careers. We already partner with 40K+ companies of all sizes, serving more than 1M+ members, and have reached €800M+ in ARR.
Prevention as the new norm. That's what we're building with our team of 1000+ people. If it speaks to you: we're hiring across France, Spain, Belgium, and Canada. And beyond.
⭐ THE ENGINEERING TEAM ⭐
In our engineering team, we build the infrastructure, interfaces, and applications to provide first-class service to our members, health professionals, and even ourselves!
Being an engineer at Alan means joining a team of talented, committed and passionate engineers, with a lot of product interaction.
We move fast, with a lot of ownership, and are proud to tackle big problems!
Our process is very simple: those who make product decisions are the same ones who build them.
Want to know more? Read this article https://blog.alan.com/tech-et-produit/engineering-organisation-at-alan if you want to discover how life as an Engineer at Alan and this article https://medium.com/alan/engineering-career-path-at-alan-e5de5c2e740c on our Engineering career pat
🛠️ THE STACK
Python services on FastAPI, async SQLAlchemy and PydanticAI, Temporal for durable execution, PostgreSQL, MCP, Daytona sandboxes, various LLM model providers, Terraform on AWS, React with Vite. Our main backend is a Flask monolith. We deploy daily and believe in distributed ownership: you build it, you own it.
🤖 THE AGENTIC PLATFORM AREA
In 2027, we expect to run more AI agents than we have employees.
Agents already do real work at Alan. They answer members' health questions, resolve support requests, read medical documents, detect fraud, automate operational tasks, and write a large share of our code.
But every team building an agent ends up solving the same hard problems again: how to call models safely with health data, how to give an agent tools and the right permissions, how to know whether a prompt change made it better or worse, how to debug what it did. The result: several agent frameworks and evaluation stacks, each with its own security and compliance story.
The Agentic Platform area was created to fix that. Our mission: make it cheap, fast and safe for every Alaner or crew to build, run and improve agentic products and workflows.
We own the environment agents run in (infrastructure, runtime, evaluation, observability, skills lifecycle), with security, privacy and compliance by design, and the generalist products on top: Hopper and the Agent Studio. Product teams and all Alaners own their agents, skills, and their outcomes. We give them the paved road.
Two crews, one team:
- Hopper is every Alaner's AI companion. Five months after it started, 80% of Alaners use it, 3k sessions are created each day, and about 30% of our pull requests are opened from it (watch how we built it https://www.youtube.com/watch?v=hAu7Ngfd0ys).
- AI Product Platform builds the shared blocks every AI product at Alan runs on. For instance, our ops teams now build agents autonomously on our Ops AI Platform (read how https://medium.com/alan/how-we-bootstrapped-an-ai-agents-platform-for-operations-at-alan-fefd87ed6314, watch the talk https://www.youtube.com/watch?v=_pg1ij8k5Rw).
🧩 PROBLEMS YOU'LL ONLY FIND HERE
Run agents at scale
- 🧠 One runtime for every AI product. A shared agent framework, from a single model call to multi-agent workflows with live streaming, and one gateway to every model: routing, failover across providers, quotas and cost per use case built in.
- ⏳ Agent sessions that last for hours. Durable sessions, started from the web, Slack, Linear or GitHub, that survive pod restarts and deploys, and pick up exactly where they left off.
- 📦 Sandboxes at scale. An isolated environment for every agent session and every evaluation, on an infrastructure compliant with sensitive data, with negligible boot overhead.
Make agents trustworthy and proactive
- 🔌 One tool gateway to the whole company. An MCP gateway that connects agents to internal tools and data, external services, computer use, managing identity and credentials.
- 🔐 Agents safely acting on critical data. We give each agent no more than the access it needs. Every tool call goes through identity, permissions, confirmations and audit. No sensitive data leaks.
- 🧪 Evals you can replay in time. Re-run an agent or skill against production as it was at a past point in time, on real data, safely isolated. Anyone can ship knowing their change is an improvement rather than a regression.
- 🔁 Agents and skills that proactively fix themselves. Every run is traced, and every human correction, failed tool call or escalation becomes a prioritized fix that Hopper can apply.
Put agents in everyone's hands
- 📚 Skills as a product. A registry where Alaners and teams publish, share and version their skills, so that nothing needs to be repeated twice.
- ✨ AI that non-engineers can own. Alaners who ship code through Hopper, Operators or Sales who build their own agents or skills in our Studio and operate them from their day-to-day tools.
- ⏰ Automations and artifacts. Agents that run on a schedule or on an event, and build the dash