The role
Our agents read dated public sources, spot the accounts that have a reason to buy, find the person who decides, and draft the first message. The hard part isn’t the demo: it’s quality, week after week, across entire markets.
You’ll work directly with the founder, on a team where agents already write part of the code. Your job: design, measure, and harden our agents, from data collection to a message that’s ready to send.
What you’ll do
- Design and improve the agents that spot buying signals, qualify accounts, and write messages.
- Build the evals that tell us whether an agent is getting better: test sets, traces, and human review.
- Make the data pipelines dependable: collection from public sources, deduplication, and a source and a date for every fact.
- Pick the right model for each task and keep its cost and latency under control.
- Ship to production every week, then watch what actually happens for customers.
Our stack
- TypeScript end to end: Bun, React, Hono, PostgreSQL with Prisma.
- Agents and jobs orchestrated with Trigger.dev; models called through OpenRouter and the AI SDK; traces and evals in Langfuse.
- Linear for tracking, Notion for docs, and coding agents in daily use.
Your first 90 days
- Day 30: you’ve shipped to production and followed an agent’s traces end to end.
- Day 60: you own an agent or a data pipeline, along with its quality metrics.
- Day 90: you’ve moved a quality metric a customer cares about, and you’ve proposed what comes next.
About you
- Two to five years of software engineering, some of it with language models in production.
- You measure before you claim: an eval beats a hunch.
- TypeScript or another typed language, SQL, and experience with unreliable third-party APIs.
- A real interest in understanding your users’ work: here, salespeople.
What you don’t need
- A PhD or a machine learning research background.
- Model training experience: we use models, we don’t train them.
- Sales experience: you’ll learn it with us.
What we offer
- A salary of €50,000 to €65,000 gross per year, depending on experience.
- Real ownership of a product customers use every week.
- Direct access to the founder and to product decisions.
- A desk at Station F, in Paris’s 13th arrondissement.
How to apply
Email us with the “Apply” button: your résumé or LinkedIn profile, plus a few lines on what draws you to this role. No formal cover letter.
Then four steps: a 30-minute call with Mayeul, a practical exercise of two hours at most that you present to the team, an in-person meeting at Station F, then a written offer. We get back to you within five business days at every step.
We consider every application regardless of origin, gender, age, disability, or background. Tell us if an accommodation would help, and we’ll arrange it.