October 9, 2024
Canadian fintech startup Evolving Intelligence recently launched a groundbreaking AI-powered private wealth management platform designed to track wealth transfer risks, provide up-to-date insights on regional tax, compliance, and other rules, and automate tasks.
Financial advisors face a daunting challenge in transferring their clients’ assets around the world. Every move requires expert knowledge of regional rules covering estates, taxation, regulatory compliance, documentation, and much more.
“Advisors often function in silos and, faced with a complex scenario, may not even know what to ask. By creating a private, longitudinal layer of knowledge, our AI platform provides them with a holistic picture and puts them in control.”
Ali Saleh
President, Evolving Intelligence
Built with the help of experts from the Vector Institute, Evolving Intelligence trained their AI models on domain expert knowledge and up-to-date regional information, ensuring the platform provides financial advisors with the AI equivalent of a private team of regional experts and assistants. Evolving Intelligence came to Vector through the FastLane program, signing up for the Data Readiness, Model Development, and Model Deployment (DaRMoD) program last summer. The Ontario-based company had already collected some internal data, but wanted to accelerate its preparation. That experience led their team to join Vector’s concurrent Recommender Systems Bootcamp to develop their system’s capabilities. The work that the Evolving Intelligence team carried out through the bootcamp allowed them to sharpen their use case and demonstrate the incredible potential of their platform.
Extra care was taken to meet their high requirements for data privacy. One of Vector’s applied machine learning scientists and a project manager from Vector’s applied AI projects team served as mentors and technical leads, working one-on-one with company teams on their use cases over three months, including a three-day bootcamp. ”We attended the Recommender Systems Bootcamp to learn the best methods for building AI processes that can quickly and reliably identify financial transfer risks,” says Saleh. exposed us to lots of different types of models. We learned a lot — beyond our expectations.”
Participating in the bootcamp enabled the Evolving Intelligence team to test the relative advantages of different models and approaches and build a functioning proof of concept against which they could check different solutions. Working with the guidance of Vector applied machine learning scientist, Responsible AI, Shaina Raza, PhD, the team broke their challenge into two distinct problems: (1) categorizing the risk of transferring a given asset within or across jurisdictions, and (2) recommending appropriate tasks based on the specific risk profile of a given asset transfer.
To categorize the risks in different asset transfer scenarios, the team needed a model that could handle a multitude of factors, including financial regulations, tax laws, political stability, currency exchange rates, and market conditions, which vary significantly over time and across jurisdictions. Working in collaboration with Raza, the team selected a model designed to handle a large number of input features and complex interactions. For their proof of concept, they incorporated several categories, including the type of transfer, the type of asset, and the source and destination jurisdictions. With their risk categories established, the team incorporated a mix of methods to recommend the relevant tasks for every asset transfer risk profile.
“The Evolving Intelligence team came in with a very good strategy and showed great interest and curiosity in learning as well as implementing different recommendation techniques to accomplish their unique use case. They brought a lot to the table and made the most out of the opportunity to work together.”
Shaina Raza
Applied Machine Learning Scientist, Responsible AI, Vector Institute
Since participating in the Recommender Systems Bootcamp, Evolving Intelligence has continued to refine its system and add capabilities. Saleh reports that their team has effectively tripled the size of their data set, fine-tuned their model, and incorporated additional features, such as increasing the number of risk categories from seven to 12. They deployed their model on their platform and it is already garnering the interest of prospective clients. “It has been great working with Vector,” says Saleh. “The mentorship and support we received was tremendous, including the office hours. We organized ourselves to make the most of this valuable resource.” Capitalizing on their momentum, Saleh and his team have returned to Vector for a third program as they sharpen their skills and further build out their platform.
As Evolving Intelligence continues to refine its platform, its success showcases Vector’s broader mission of driving responsible AI innovation and fostering sustainable growth in Canada’s AI ecosystem. As a trusted partner in building, developing, and adopting AI, Vector continues to bridge the gap between cutting-edge research and real-world applications. Through programs like FastLane and collaborations with companies like Evolving Intelligence, Vector not only advances AI capabilities but also ensures that these advancements are implemented responsibly and ethically. As we look into the future, Vector remains dedicated to translating our research leadership into tangible benefits for industry partners, ultimately helping Canada win with AI on the global stage.
For more on the use of recommender systems, check out the paper, co-authored by Vector researchers, “A Comprehensive Review of Recommender Systems: Transitioning from Theory to Practice.”
