Why we partnered with Mila

There's a persistent gap between what AI can do in research papers and what it actually does in production. That gap is especially wide in healthcare and other regulated industries, where the stakes of getting it wrong are measured in patient outcomes, not just user metrics.

We partnered with Mila – Quebec Artificial Intelligence Institute because closing that gap takes more than reading the papers: it takes proximity to the people producing the science, and exchange that runs in both directions. I've wanted this for a while.

Mila is the world's largest academic AI research centre specialized in deep learning, founded by Professor Yoshua Bengio and home to a community of over 1,500 members. Its researchers publish across the areas that matter most to our healthcare work: medical imaging, drug discovery, clinical decision support, and health data analysis.

Mila is also a big part of why Montreal matters in AI at all: the talent it trains, the research it anchors, the global labs it helped draw here. For a Montreal company, it's the epicentre of the ecosystem we build in.

Our first collaboration is already underway: a Mila PhD student is joining Toboggan Labs this summer, working alongside the developers, data scientists, and designers who build and ship our healthcare AI systems. That's the exchange already running in both directions.

Over the longer term, the partnership gives our team access to professional development and high-level knowledge-sharing inside Mila's ecosystem: a structural way to stay close to advances in deep learning, AI safety, and healthcare AI as they emerge.

Our own work lives on the production side of that divide, and we write about it regularly: what it takes to run production-ready AI agents in healthcare, where bounded autonomy and audit trails count for more than impressive demos, and how to build evaluation systems that hold up when domain-expert time is scarce. Turning research into systems a clinician can rely on is a craft of its own.

That work spans the full arc: advising on technology strategy and AI governance, designing products clinicians actually adopt, building the applications and the data science behind them, and running them in production under healthcare's regulatory constraints. All of it serves the same goal: helping clients understand not just what to build today, but how the shifting AI landscape affects their technology strategy for the years ahead.

What's next?

The partnership strengthens our capabilities. It doesn't change our philosophy. We still believe that problem understanding should precede solution design. We still prioritize strong UX and human-in-the-loop systems over fully automated ones in high-stakes domains. We still invest in evaluation frameworks that let us measure whether our systems actually work before we deploy them. What changes is the depth of scientific support behind that approach. We'll write about what we learn.