As AI4Good Lab prepares to launch its tenth cohort, Mila sat down with its co-founders, Angelique Mannella, Vehicle Technology Transformation Lead at AWS, and Doina Precup, Research Director at Google DeepMind and Canada CIFAR AI Chair at McGill University and Mila, to reflect on a decade of training women and gender-diverse participants in applied AI, and to talk about what comes next.
Ten years ago, a small group of 25 participants gathered at McGill University for what would become one of Canada's longest-running machine learning bootcamps for women and gender-diverse people. The idea behind AI4Good Lab was equal parts practical and visionary. It set out to create a space where people who had historically been left out of AI could not only learn the technical skills, but also see themselves as shapers of the technology. As the lab prepares to welcome its tenth cohort, its co-founders look back on a journey that neither of them could have fully anticipated when they first started brainstorming in 2016.
A Problem Worth Solving
For Angelique Mannella, the motivation was personal. Having worked in the tech industry for many years, she had thought carefully about what had been missing from her own experience as an engineering student.
"Fast forward many years later having acquired different experiences and resources, what could I create that could meet the needs of helping expand personal networks, providing opportunities for growth and learning outside the confines of the traditional academic environments," she says.
For Doina Precup, a researcher and professor whose path took her from Romania to Canada, the motivation came from a different angle. What struck her most was the stark contrast of the gender gap she encountered after arriving in North America.
"I grew up in Romania and there wasn't a gender gap particularly in math and science or engineering," she explains. "My mom was a computer scientist, just like my dad. So when I moved to the US and then to Canada, I found it strange."
What began as a cultural puzzle became a recognition of structural barriers. The lab's signature idea of using social good problems as the thematic anchors for team projects was born from a conviction that participants would be more motivated if the work felt meaningful from the start.
Designed to Adapt
Both founders are quick to note that the name was chosen deliberately. The lab was always an experiment in how to teach machine learning to a more diverse audience, and the founders treated it as such.
"We thought not just that students would be experimenting with ideas," says Precup, "but that we ourselves would be experimenting with different ways of teaching and creating these projects."
The curriculum has evolved considerably, from deep learning and reinforcement learning toward the generative AI landscape of today. The geographic footprint has grown from that original Montreal cohort to four cohorts running across Montreal, Toronto, Edmonton and online. Rather than setting up a separate nonprofit, the founders chose to work within a network of complementary partners — McGill, CIFAR, OSMO, then Mila, Vector and Amii — sharing resources to build a pan-Canadian presence.
"Really work with the community, see where we can collaborate or share resources to accomplish the mission versus doing it all on our own," says Mannella. "That has been part of the ethos."
What Ten Years of Alumni Has Shown
Ask either founder what the lab has accomplished and they return to the people who passed through it. Mannella points to three things she hears consistently from alumni: confidence in their technical skills and their place in the field, access to training that academic institutions are often too slow to provide, and the peer network.
"The alumni networks that have formed, how that peer network can be a source of learning, a source of friendship, a source of encouragement… that's something we receive feedback on as being pretty powerful," she says.
Precup describes a compounding effect that is hard to measure but intuitive. Alumni have gone on to diversify their own teams, attracting more talent and shifting what AI development looks like in Canada from the inside.
Progress, and the Work That Remains
Neither founder is ready to declare victory on diversity in Canadian AI. Precup points to a troubling pattern in the data. While undergraduate computer science programs have reached reasonable gender parity, those numbers drop sharply at the graduate level, and the trend has worsened since the AI boom. Mannella acknowledges the shifting political climate around diversity and inclusion without letting it alter her sense of the mission.
"The need is still there, the impact is still there," she says. "It doesn't take away from the mission of the lab or the need to keep going."
Looking Ahead
For the tenth cohort and beyond, Mannella's priority is deepening the Canadian footprint and reaching communities not yet well served by the program. "I’m very proud of what we’ve accomplished over the last 10 years and the footprint we’ve had," she says. "But I think there’s definitely an opportunity to go bigger and broader."
Precup looks toward questions the lab has not yet resolved, among them whether to reach younger students, whether to broaden its scope beyond gender diversity, and whether the model could eventually be exported internationally. Her instinct is measured, though. Growth is only meaningful if quality and accessibility can be preserved.
What both founders agree on is that the program's longevity comes from its willingness to keep learning. The curriculum, the delivery format, the partnerships and the geographic reach have all changed. What has stayed constant is the belief that brought the lab into existence in the first place. AI will be shaped by those who build it, and the people building it should reflect the full diversity of the lives it touches.
The tenth cohort is about to find that out for itself. Visit the AI4Good Lab website to learn more about this groundbreaking initiative, and check out this recent article to hear from three alumni about their experience with the program.
