LLMOps Engineer
Operates large language models in production: evaluation, inference cost, observability, RAG and reliability of LLM-based systems.
Why this hire matters.
Getting an LLM into a demo takes a day; keeping it in production is a profession. Between demo and product lies everything the LLMOps engineer masters: systematic evaluation (without which every prompt or model change is an invisible regression), inference cost control (which can destroy your margins), latency, agent observability, guardrails against hallucinations and injections. It's the best-valued AI profile on the French market — logically: the difference between a reliable AI product and an unpredictable gadget runs through them. If your product rests on LLMs and nobody owns evaluation and cost, you're flying blind on your most strategic feature. Young role: judge on systems built, not years of tenure.
Key missions.
- Build LLM evaluation infrastructure (evals, internal benchmarks)
- Optimize inference cost and latency (routing, caching, quantization)
- Set up observability for LLM systems and agents (traces, feedback)
- Industrialize RAG, fine-tuning and prompt management
- Deploy guardrails (hallucinations, safety, injections)
- Version prompts, models and datasets
- Arbitrate model choices (API vs open-weights, specialization)
Skills.
Technical skills
- LLM evaluation (LLM-as-judge, eval datasets)
- Serving and inference optimization (vLLM, TGI)
- RAG and vector databases
- LLM observability (LangSmith, Langfuse)
- Production Python
Expected qualities
- Experimental rigor
- Cost awareness
- Intensive tech watch
Common stack
Salaries 2025-2026
| Level | Experience | Annual gross base |
|---|---|---|
| Junior | 0-2 yrs | 55–75 k€ |
| Mid-level | 2-5 yrs | 75–95 k€ |
| Senior | 5-8 yrs | 90–115 k€ |
| Lead / Staff | 8+ yrs | 115–160 k€ |
Paris market ranges, 2025-2026.
Outside the Paris region, expect 10 to 20 % less.
Sources : Silkhom Baromètre IA 2026 (moyenne 85 400 €) · Factoriel 2026 · Get in Talent 2026
Hiring this profile.
Typical background
MLOps or ML engineer specialized in LLMs, or a senior developer who has built GenAI systems in production. Real systems are the proof.
When to hire
As soon as LLMs are core to the product and costs, regressions or incidents become visible — very early for AI-natives, Series A/B otherwise.
Career path
Senior → Lead LLMOps / Head of AI Platform → Chief Data & AI Officer.