Machine Learning Engineer - Reinforcement Learning
Blue Yonder · Paris
About the AI Studio
The AI Studio's mission is to find the fastest possible path to an autonomous supply chain. We're developing AI agents, learning systems, training models, and more to overcome the biggest challenges remaining in the global supply chain.
In short, we are having a lot of fun.
Your Mission In This Role
We’re looking for an ambitious ML Engineer focused on LLMs, agents, and reinforcement learning to help build the training, evaluation, and tooling systems behind robust AI decision-making products.
You’ll work across LLM fine-tuning, agent environments, reward modeling, evaluations, data pipelines, and AI workflow tooling. The role is hands-on: designing experiments, shipping production code, improving model behaviour, and building the infrastructure that lets us learn quickly from both automated and human feedback.
You’ll help shape how we use LLMs inside agentic systems, how we evaluate model and agent performance, and how we turn feedback into better training data and better behaviour.
This role requires mandatory RL training experience with LLMs, including designing and iterating on rewards, reviewing LLM traces, identifying reward hacking or shortcut behaviour, and understanding when the reward signal, environment, or training process needs to change.
Responsibilities:
- Design and implement LLM-powered agent environments for supply chain decision-making
- Fine-tune, adapt, and evaluate LLMs for domain-specific reasoning and decision support
- Design, test, and iterate on reward functions that capture the behaviors we want from LLM agents
- Review LLM traces and rollouts to understand model reasoning, failure modes, reward hacking, and shortcut behaviour
- Identify when an LLM is exploiting the reward function, escaping the intended RL process, or optimizing for proxy metrics instead of the real objective
- Improve reward models, environment design, prompts, tools, and feedback loops based on observed model behaviour
- Build evaluation frameworks to measure model quality, agent performance, robustness, and failure modes
- Create data pipelines for training, fine-tuning, preference data, synthetic data generation, and human feedback collection
- Develop tooling that improves how the team builds, tests, debugs, and deploys AI-assisted workflows
- Experiment with RL, RLHF, RLAIF, reward shaping, policy optimization, and agent training techniques
- Document what works, what fails, and why, so we can compound our learnings over time
- Stay close to the frontier of LLMs, agents, evaluations, and applied AI engineering
We want to talk if you:
- You've trained or fine-tuned LLMs
- Are excited about AI-assisted tools and getting the most out of them
- Build & customize your own AI workflows
- Have experience working with AI agents and RL environments in production
- Are proficient in Python and PyTorch
- Can balance research exploration with shipping working code
- Hands on experience with RL techniques (reward shaping, policy optimization, RLHF)
- Thrive in fast-moving environments where priorities shift
- Care about craft in your work
- Are curious about why things work, not just that they work
Bonus points if:
- You have experience with human-in-the-loop ML systems
- You've built evaluation frameworks for open-ended tasks
- You're familiar with supply chain, logistics, or operations domains
- You have a side project that shows you can't stop tinkering
#LI-HG1
Our Values
If you want to know the heart of a company, take a look at their values. Ours unite us. They are what drive our success – and the success of our customers. Does your heart beat like ours? Find out here: Core Values
All qualified applicants will receive consideration for employment without regard to race, color, religion, sex, sexual orientation, gender identity, national origin, disability or protected veteran status.