AI Engineer
Pubblicato il 09-09-2026 - Mia-Platform in Milano
Job description
The role
We are looking for an AI Engineer, to join our Mia-Platform group. Specifically, the person will work on our Mia-Care Team.
As an AI Engineer, you'll be at the core of what makes Mia-Care tick. Our platform uses LLMs to analyze medical device software, identify compliance gaps, and suggest actionable remediations. You'll own the intelligence layer that makes this possible, from model selection and adaptation to production deployment and cost optimization.
This isn't a role where you plug in an API and call it a day. You'll be shaping how AI works inside a highly specialized, high-stakes domain, where quality and reliability are non-negotiable.
Responsibilities
- Select, adapt, and evaluate pre-trained LLMs for domain-specific tasks, including prompt engineering, RAG pipeline design, and targeted fine-tuning where justified.
- Design and validate prompt pipelines using structured evaluation frameworks: ground-truth test sets, automated metrics, and regression suites, with documented acceptance criteria for production deployment.
- Own the full MLOps lifecycle: model versioning, monitoring, evaluation pipelines, and continuous improvement.
- Optimize for cost and scalability. You'll obsess over inference efficiency, latency, and making sure what you build can grow without burning the budget.
- Build AI pipelines that are reliable, observable, and production-grade from day one.
- Collaborate directly with product and domain experts to translate regulatory complexity into AI-solvable problems. No translation layer, no ticket queue.
How we work
- Full-stack AI mindset. You own your work end-to-end, from research and prototyping to deployment and monitoring. We value builders who take complete responsibility for their features without hand-off friction.
- Practical Engineering. We are a team of builders, not theorists. We prioritize writing clean, maintainable code and automating repetitive tasks to ensure our infrastructure remains robust and