> Markdown version of [/jobs/ext/4004275-ai-software-engineer-f-h](https://www.wearedevelopers.com/jobs/ext/4004275-ai-software-engineer-f-h). Every page supports `.md` or `Accept: text/markdown`. Links point to the HTML versions so they work for humans too. Agent guide: [/agents.md](https://www.wearedevelopers.com/agents.md).

---

# AI Software Engineer (F/H)

- **Company:** Fabriq
- **Location:** Paris, France
- **Experience:** Experienced
- **Salary:** €70,000.0 - €80,000.0
- **Contract:** Permanent contract
- **Skills:** LangGraph Framework, Artificial Intelligence, Amazon Web Services, Code Review, Octave, Software Engineering, TypeScript, Backend, Agentic-AI, Vue.js, Kubernetes, Sentry, Front End Software Development, Terraform
- **Published:** October 29, 2026
- **Apply:** https://www.welcometothejungle.com/en/companies/fabriq/jobs/ai-software-engineer-f-h_paris

## About the Role

Must-haves :

 * Software engineering depth: 4 to 5 years minimum. Real craftsmanship, clean design decisions, and the ability to explain the architecture trade-offs you made and why.
 * Daily hands-on use of AI to build software. Having never used AI to develop is a deal-breaker on this role.
 * Comfort with ambiguity. You move forward, decide and ship when nothing is specified in advance, and you adapt when the roadmap shifts.
 * Builder mindset with evidence. Concrete examples of things you built from zero, prototyped, tested and iterated on.

Nice-to-haves (optional) :

 * A first hands-on experience with agentic systems. Expertise is welcome when it is real : a strong engineer with genuine curiosity and early agentic work beats a self-declared two-year agentic expert who cannot explain concretely what they built.
 * A data background. The testing logic differs from pure software: working on datasets and analysing where a system succeeds and where it fails is a plus, and it can be learned here.
 * Collaborative posture. the team will evangelise AI across the other product teams. Showing and explaining your work is expected.

Stack :

Backend: TypeScript on Deno. Frontend: Vue.js. Infrastructure: AWS with EKS, Terraform and CDKTF, plus Kubernetes for on-prem customer deployments. Observability: Honeycomb and Sentry. On the AI side: frontier models called through AWS Bedrock, with everything agentic hand-built and no framework on top.

 4. Package and working conditions
 * Salary: €70K to €80K fixed, depending on experience.
 * Location: Paris office at 18 rue Yves Toudic, 75010. Presence matters most during onboarding and for team dynamics. Remote is possible, Paris is preferred.
 * Working language: French & English, fluency required.

## Description

Context : You join the AI team, a two-person unit led by Jean (Product Lead) alongside Halim, Senior Engineer who has built much of the Fabriq product. This team operates as a prototyping squad at the front edge of the company: it tests new ideas, measures market appetite and decides whether Fabriq goes there. Products are in alpha, cycles are short, the roadmap is directional. The team also works weekly with Octave, co-founder, which gives you direct exposure to the strategic decisions.

Objective : Become Halim's pair and second engineer on AI, reach full autonomy on the stack and the customer context within 6 months, and take Fabriq's AI products from alpha to something industrial teams rely on every day.

Challenges :

 * Taking AI prototypes to production quality. Conversational and voice-driven agents, plugged into real industrial data and used by real teams on the floor. You take them from alpha to something people rely on daily: answer quality, reliability, and the cost per interaction that decides whether a use case scales or gets dropped.
 * Prototyping to convince. Part of the roadmap is co-built with partner customers, which means it is sold before it is built. Specs move as customers shape them, timelines are tight, and you build fast to prove a use case is worth pursuing. Testing a lot and abandoning a lot is part of the deal.
 * Agentic engineering from scratch. No Mastra, no LangGraph, no off-the-shelf framework. Everything is hand-built in TypeScript and Deno on top of frontier models called through AWS Bedrock. When the team needs something like agent memory compaction, you design it yourself. This is genuinely custom, sparsely documented territory, and that is precisely the technical interest of the role.
 * Setting the technical direction with Halim. You are the second engineer, not an extra pair of hands: code reviews, architecture debates, and the freedom to arrive with your own proposals rather than waiting for a ticket.

Your first 3 months are dedicated to ramping up on the Fabriq stack and product, contributing to the AI features already in the hands of beta customers, and shipping your first POCs.