Hybrid Prompt Engineer

at Nabla

• Job title: Prompt Engineer
• Job type: Contract
• Experience level: Mid level, Senior
• Salary: Not specified
• Degree requirement: No Education Requirement
• Tech skills: Python
• Location requirements: Hybrid, Paris, France

About the role

• First Prompt Engineer at Nabla responsible for quality of LLM systems and prompt design. Collaborating with Product, ML, and Clinical teams to streamline clinical workflow through AI.

Responsibilities

• Design, write, and maintain high-quality prompts for: Clinical note generation
• Evaluation / “judge” models assessing hallucinations, recall, structure, tone, and adherence to clinical standards
• Systematically iterate on prompts based on failure modes, edge cases, and real-world usage.
• Define clear, measurable quality criteria for each type of output, including accuracy, completeness, style, and clinical relevance.
• Build and maintain prompt-based evaluation frameworks to consistently score and compare outputs across models and prompt versions.
• Identify blind spots, regressions, and trade-offs in model behavior.
• Treat prompts as production artifacts: versioned, documented, tested against known failure cases
• Create reusable prompt patterns and guidelines to ensure consistency and maintainability.
• Work closely with ML engineers to surface model limitations and propose prompt-level mitigations.
• Partner with Product and Clinical teams to translate qualitative expectations into explicit, testable instructions.
• Act as a bridge between “what clinicians expect” and “what LLMs actually do.”

Requirements

• Prior hands-on experience with prompt engineering for production LLM systems (not just experimentation or demos).
• Strong written communication skills and an exceptional attention to detail.
• A rigorous mindset: you enjoy defining criteria, edge cases, and evaluation frameworks.
• Comfort working with:
• LLM APIs (OpenAI, Anthropic, Gemini, etc.)
• Structured prompt formats (system / developer / user messages)
• Basic scripting or tooling to run prompt tests (Python, notebooks, or similar)
• High tolerance — and even enjoyment — for iterative, detail-oriented work that requires patience and taste.
• Ability to reason about hallucinations, recall failures, ambiguity, and instruction-following limitations.
• Interest in healthcare, clinical workflows, or safety-critical AI systems is a strong plus.
• Autonomous and self-directed; comfortable operating with minimal process in a fast-moving startup environment.

Benefits

• Public transportation costs covered at 50%
• Exercise class during the workday: Yoga, running, pilates, HIIT
• Culture of trust & accountability — your output matters more than your clock-in time