🛠 Tech Stack

💼 About This Role

You'll own our inference infrastructure end-to-end, optimizing latency, throughput, and cost across our model fleet. You'll build and scale model serving with TensorZero, vLLM/SGlang/TRT, and Kubernetes, and turn research into production-ready products. This role bridges Research and Product to ship models that are fast, cheap, and production-ready.

🎯 What You'll Do

• Own inference infrastructure end-to-end: optimize latency, throughput, and cost.
• Build and scale model serving with TensorZero, vLLM, SGlang, TRT, and Kubernetes.
• Design and maintain vector search pipelines with vector storages.
• Turn research into product: format, sample, deploy experimental models.

📋 Requirements

• 3+ years shipping high performance ML systems in production.
• Deep hands-on experience with inference optimization.
• Comfortable across the stack: from CUDA kernels to Kubernetes manifests to Grafana dashboards.

✨ Nice to Have

• Rust experience
• Custom Triton kernels
• Benchmarks

🎁 Benefits & Perks

• 💰 Competitive salary with equity
• 🏖️ 20 days of paid vacation
• 🏢 Hybrid work in Paris + relocation package
• 🏥 Best medical insurance in France
• 🖥️ All hardware, tools, and AI subscriptions covered

📨 Hiring Process

Estimated timeline: 1-2 weeks

1. 1Intro call with a colleague· 30 min
2. 2Take-home exercise· 2-3 hours
3. 3Technical interview· 1 hour
4. 4Final call with CEO and CTO· 45 min

🚩 Heads Up

• Department listed as 'Research team' but role is MLE/MLOps mixing research and production.
• The description includes 'Familiarity with support metrics' which is not typical for MLE role.