RL Research Engineer
Enregistrez cette offre et organisez votre recherche
Créez un compte gratuit pour enregistrer des offres d'emploi, créer des alertes et revenir à cette liste depuis votre tableau de bord.
En continuant, vous acceptez nos Conditions d’utilisation & Politique de confidentialité.
Hybrid | Paris / London / Amsterdam | €150,000 + Founding Equity
The best AI engineers aren't waiting for a job ad. They're waiting for the right moment. This might be it.
Anthropic leadership has discussed spending over $1 billion on RL environments in the next year alone. The market is moving fast and at scale. One pre-seed company just closed a round at a valuation most Series A companies would envy, with frontier AI labs already as paying clients and acquisition interest already on the table (x2).
The reason? They cracked something the rest of the market hasn't.
Most teams build RL environments from synthetic data - easy to demo, easy to commoditize, brittle. This team mines real human behavioural data - how domain experts actually reason, decide, and solve complex tasks over long-horizon workflows. 10 -1000+ step environments. Closed-loop systems where environments, data, training, and evaluation are tightly integrated. Not proxies. Not short cuts.
25-30% uplift in model task success rates. 50-65% more training signals. Evals that reflect how humans actually work.
The founding team is being built right now. Are you going to be in one of them?
- €150,000 + founding equity at a ~$40M seed valuation
- Environment design treated as a first-class problem
- RL for real agents - tool-use and genuine agentic systems, not pipelines
- London / Paris / Amsterdam hybrid working
- Get in now - before this role disappears and you're watching from the outside
On the LLM post-training side, you'll have:
- Hands-on experience post-training foundation models with RL - GRPO, DPO, RLVR or similar
- Worked on smaller open-weight models (not just closed APIs or sheer scale)
- Owned the full post-training loop - data, rewards, training runs, evaluation
You’ll love this if...
- You've trained agents with RL and you know the difference between that and prompting an agent
- You've built the evaluation-to-training loop - environments, rollouts, reward signals feeding post-training
- You think turning real human expertise into training signal is one of the most underrated problems in AI right now
- Your RL is the genuinely hard kind - non-trivial reward design, long-horizon, multi-agent
- Bonus if your RL experience is in a real expert domain - CAD / CAE / mechanical engineering, aerospace, finance, insurance or law
- You want to work at the forefront of innovation with frontier labs
Roles like this don't stay open. If it fits - apply today.
Because great teams are Built Different.