Skip to content

An AI assistant for clinicians

Client
A Québec healthcare institution
1 min read
A stethoscope in front of a laptop on which a healthcare professional is typing

Deploying a conversational assistant that helps clinical teams find internal protocols and procedures in seconds, within a strict framework for protecting health information.

Context

Care teams at this institution spend a significant share of their time searching for scattered information: clinical protocols, administrative procedures, internal memos, and practice guidelines. Spread across multiple intranets and shared drives, these documents are often hard to find when every minute counts.

Leadership wants to leverage generative AI to ease this burden, provided the solution complies with Law 25, guarantees that data is hosted in Canada, and integrates with the tools staff use every day.

The challenge

A general-purpose assistant is not enough: answers must be reliable, sourced, and up to date, and must never fabricate information. Even the slightest deviation on a clinical protocol can directly affect the quality and safety of care.

The project must also win over demanding users, secure approval from governance and security bodies, and build in, from the design stage, continuous evaluation of answer quality and control of inference costs.

Technologies

  • vLLM
  • Mistral (open-weight models)
  • LangGraph
  • PostgreSQL / pgvector
  • MLflow
  • Kubernetes
  • Microsoft Entra ID

What do you want to build?

Let’s talk