Senior Prompt Engineer (AI Clinical Products)

As a Senior Prompt Designer, you will play a defining role in shaping the quality, reliability, and evolution of MEDFAR's generative AI features — most notably Coeur Way, our AI-powered clinical scribe, and emerging AI capabilities within MYLE EMR. Reporting to the Director of UX, you will sit at the intersection of product, clinical quality, and AI craft, working closely with product designers, developers, QA, and client-facing teams to ensure that every AI-generated output meets the high standards required in a healthcare setting.

This is a senior individual contributor role with a clear growth path toward mentoring and leading a prompt engineering practice as MEDFAR's AI product surface expands.

Main Responsibilities

• Audit and re-architect Coeur Way's existing monolithic prompts into a modular chain of focused, maintainable prompts — improving robustness, debuggability, and adaptability.
• Lead the selection, evaluation, and pairing of language models to specific prompt tasks, fostering a model-agnostic approach that reduces vendor dependency and optimizes for quality and cost.
• Own the quality of AI-generated clinical outputs: define what "good" looks like, identify failure modes, and drive continuous improvement through structured validation processes.
• Evolve and re-align our existing evaluation framework, currently covering 7 clinical note quality rubrics, improving its coverage, automation potential, and alignment with clinical ground truth.
• Establish and maintain a systematic approach to output validation, including edge case identification, regression testing for prompt changes, and documentation of known limitations.
• Design and implement new prompts and evaluation suites to support new features in Coeur Way and MYLE EMR.
• Collaborate with the GenAIOps to ensure prompts are production-ready — observable, versioned, and deployable through established CI/CD processes.
• Develop a deep understanding of clinical workflows and the needs of healthcare providers, maintaining close relationships with client-facing teams and end users.
• Stay current on the rapidly evolving LLM landscape — new models, prompting techniques, evaluation methodologies — and bring relevant findings back to the team.
• Contribute to internal documentation, standards, and best practices for prompt engineering at MEDFAR.
• Mentor future prompt designers as the team grows, and participate in hiring decisions as the practice scales.

Working conditions:

• Contract: Permanent, full time (40h/week)
• Working mode: Hybrid or remote
• Occasional in-office presence may be required during the year (for events or team meetings, for example).
• Candidates must reside in the province of Quebec.

Qualifications

Contribute with your strengths:

• 5+ years of professional experience in AI/ML, product development, computational linguistics, or a closely related field.
• 2+ years of hands-on prompt engineering experience in production LLM systems, ideally in a B2B SaaS or regulated industry context.
• Demonstrable experience decomposing complex prompts into modular, chained architectures (e.g., using patterns such as prompt chaining, routing, reflection, or tool use).
• Experience designing or significantly improving LLM evaluation frameworks — rubric design, inter-rater alignment, and failure mode analysis.
• Strong familiarity with frontier commercial LLM APIs (OpenAI, Anthropic, Google, etc.) and the practical tradeoffs between them.
• Experience working within or alongside healthcare, clinical, or life sciences domains is a strong asset.
• Proficiency in Python or another scripting language for building eval pipelines or prompt tooling is an asset.
• Experience with self-hosted or open-weight models is an asset.
• Advanced proficiency in English, both written and spoken;
• French is also required due to frequent client interactions

Who you are:

• You understand that great prompt engineering is as much about thinking clearly as it is about technical execution. You break complex problems into focused, well-scoped sub-problems.
• You are model-agnostic in mindset: you care about output quality and understand that the best tool for a task may not be the most popular or most expensive one.