Empowering organizations through applied AI solutions
The separation between business applications and academic AI research can sometimes make it difficult for organizations to identify and adopt the right technologies to tackle specific problems.
To bridge this gap, Mila’s Applied Machine Learning Research Team works with organizations on high-impact and challenging machine learning projects. The team's unique collaborative approach aims at designing and implementing cutting-edge solutions that benefit organizations by enriching their machine learning capabilities. Whether your project requires natural language processing, computer vision, or other forms of expertise, our team's diverse skill set can help you achieve your goals.Contact us Meet the team
From the earliest stages of AI projects, our specialists aim to establish proofs of concept to help organizations capture the full value of machine learning. Our teams strive to fully understand the problem being solved and how a machine learning model will be used in production. The success of our approach is due in large part to a solid understanding of the datasets we are entrusted with and rigorous experimental protocols set up according to project objectives.
Thorough analysis of the data and rigorous experimental protocols
Industry friendly intellectual property agreements
Committed to delivering value and rigorous results
Custom approach to complex projects
Collaborative approach geared for knowledge transfer
Industry-level code and clear documentation
CDPQ collaborates with Mila to explore machine learning-based investment strategies.
We collaborated with CDPQ on a research project to explore machine learning-based investment strategies for predicting allocations of a portfolio of stocks, based solely on the past prices of those assets. Those strategies, referred to as momentum-based, showed interesting results in the recent literature and we wanted to verify the irapplicability to the investing context of CDPQ. In extension of this line of research, we tested the approach to a universe composed of S&P500 stocks rather than futures contracts and introduced transaction costs explicitly. To account for the characteristics of financial data, we applied a walk-forward training process to re-estimate the parameters, modified the objective function to ones actually used by practitioners, namely the Sharpe and Information ratios, and tested ensemble approaches. We proposed ideas based on continual learning to try to overcome the observed limitations of those models.
Leveraging reinforcement learning to optimize telemedicine for patient care at Dialogue.
Dialogue provides a virtual healthcare platform that aims to assign a patient to the right practitioner thanks to a chatbot whose goal is to collect all relevant information about a patient’s symptoms and antecedents. The system used by Dialogue is rule-based, and is therefore difficult to extend to include new pathologies or new medical knowledge. It also asks a lot of questions and doesn’t always collect all relevant information. Mila built a new model, based on reinforcement learning, that collects much more relevant information while asking a smaller number of questions. The training of this model includes a reward shaping function whose goal is to mimic the reasoning of a senior physician.
Using deep learning models to accelerate the completion and verification of claims at American Family Insurance.
American Family Insurance is interested in using deep learning models to assist its insurance professionals by accelerating the completion and the verification of claims for its automotive insurance line of products. Those models need to handle multimodal inputs such as text, categorical data, and images. The models will be designed to predict or verify the values of attributes describing individual vehicles involved in an accident, as well as overall claim-level attributes describing properties of the accident.
Optina Diagnostics partners with Mila to accelerate the early detection of Alzheimer’s disease.
Optina Diagnostics is developing the Retinal Deep Phenotyping™ platform for detecting the cerebral amyloid status and other key biomarkers for Alzheimer’s disease (AD). Given that the retina, lining of the eye, is an extension of the brain, it can provide insights into neurological pathologies. The goal of the project is to use Optina Diagnostics’ retinal hyperspectral images to detect phenotypes associated with cerebral amyloid status via direct observation of the retina. This approach is less invasive and less costly than brain amyloid positron emission tomography (A-PET) scans.
Retinal scans thus offer the possibility to improve the availability of tests which could lead to earlier AD diagnosis, and therefore improve the quality of life of impacted patients. We are using the results of a clinical study of a few hundred patients for whom both brain A-PET scans and Optina Diagnostics’ retinal hyperspectral images are available. Physicians analyzed the A-PET scans to categorize the cerebral amyloid beta status of the patients.
With this project, Mila has to work with a relatively small number of patients compared to typical deep learning datasets, as the reference standard, A-PET scans, are both expensive and inaccessible. But the large size of retinal hyperspectral cubes (in terms of spatial and spectral resolution) open vast possibilities for our research team.
We are collaborating with Optina Diagnostics’ machine learning team to accelerate the development of deep learning models and ensure that they continue to increase the accuracy and precision of their cerebral amyloid status test.
Hydro Québec’s research center and Mila collaborate to accurately predict solar irradiance in order to better understand the implications for electrical production.
We are working with researchers at CRHQ to accurately predict the global horizontal irradiance (GHI) in Quebec and the north-eastern US from zero to six hours ahead, using images from a geostationary environmental satellite. The GHI is the total amount of shortwave solar radiation received by a horizontal surface (W/m2). Better predicting the GHI will primarily improve our understanding of the potential for electrical production using solar energy and provide additional tools to manage the electrical grid. One of the main challenges in this project is the scarcity of ground truth GHI signals, as the number of pyranometers in the territory of interest is extremely small.
Our applied research experts offer consulting services to organizations working on hands-on machine learning projects. We are committed to helping your organization achieve its goals by tailoring our expertise toward your specific needs. Our team can:
We collaborate with the National Research Council of Canada and its Industrial Research Assistance Program (IRAP) to support companies working on machine learning projects.
Visit the NRC website for more information.Discover some of our IRAP Partners
« We already had some in-house expertise in AI. The collaboration with the experts of the applied research group allowed us to give a concrete boost to our current project. We were able to identify development paths and validate them in an exceptionally short time. We acquired expertise that would otherwise have taken us much longer to obtain. This interaction with Mila had a significant impact on our timeline. »
« Mila’s experts played an instrumental role in helping us advance our team’s machine learning capabilities and building internal expertise. Collaborating with Mila allowed us to build, test, optimize and deploy Roxi, the only AI algorithm for concrete performance prediction, analysis, and optimization. The algorithm is currently used by practitioners, on a daily basis, to analyze millions of cubic yards of concrete in thousands of projects worldwide. »
« Mila was instrumental in identifying and integrating deep learning models of the “transformers” type that were more efficient than the models we were previously using. At the time, these models had just appeared in the literature and had not yet been popularized. »
« Our collaboration with Mila’s experts was essential to the development of our solution. More than just an application of their expertise, their approach based on knowledge transfer allowed us to really take ownership of the solution we developed together. »