Prompt Engineer
Optimizes language-model behavior through systematic design, evaluation and iteration of prompts and system instructions.
Why this hire matters.
On an LLM product, the system prompt is code: it encodes business logic, tone and guardrails — and a bad version silently degrades every user's experience. The prompt engineer brings what 95% of teams lack: a systematic approach (evaluation datasets, measured iteration, non-regression) where others tinker on intuition. Be clear-eyed about this hire: the 'pure prompt engineer' role is being absorbed into AI engineering, and salary grids diverge sharply depending on whether you mean a prompt writer or a technical hybrid. Our read: hire the skill rather than the title — in an AI engineer, a technical UX writer or a domain expert — unless your product rests entirely on conversational quality, where a dedicated specialist is justified.
Key missions.
- Design and iterate system prompts and agent instructions
- Build evaluation datasets and measure regressions
- Optimize the quality/cost/latency trade-off of prompts
- Document and version the prompt library
- Partner with product and engineering on expected behaviors
- Test robustness (edge cases, injections, tone drift)
- Train teams on prompting best practices
Skills.
Technical skills
- Fine-grained LLM mastery (capabilities, limits, per-model styles)
- Evaluation methodology (evals, A/B)
- Technical basics (Python, APIs, JSON)
- Precise writing
Expected qualities
- Experimental rigor
- Linguistic sensitivity
- User empathy
Common stack
Salaries 2025-2026
| Level | Experience | Annual gross base |
|---|---|---|
| Junior | 0-2 yrs | 45–60 k€ |
| Mid-level | 2-5 yrs | 60–90 k€ |
| Senior | 5-8 yrs | 90–130 k€ |
| Lead (rare) | 8+ yrs | 130–200 k€ |
Paris market ranges, 2025-2026.
Outside the Paris region, expect 10 to 20 % less.
Sources : Silkhom 2026 (moyenne 49 700 €) · Glassdoor France 2026 (moyenne ~70 000 €) · Get in Talent 2026
Hiring this profile.
Typical background
Varied profiles: technical linguists, UX writers, developers, domain experts. Very recent role — judge on prompt systems actually in production.
When to hire
When conversational quality is the core of your product and prompt iteration already consumes an untooled half-time. Otherwise, embed the skill in an existing role.
Career path
→ AI Engineer, LLMOps, AI product management, or evaluation specialization at a lab.