Innovative Projects Realized

Intuitive Robot Programming for Human-Robot Collaboration with AI and AR

UI-based Evaluation System for LLM Outputs

Development of a Self-Driving Thermodynamic Laboratory

Machine Learning Applications in Psychiatric Neuroimaging

Hyperspectral Imaging – technology for real-time assessment of kidney perfusion in urologic surgery

Orbital currents: from fundamental physics to memory and AI devices

AI-based Risk Detection and Anomaly Detection

Internship in Applied Sports Medicine and Exercise Science

Adaptive EMG-Driven Control for Multi-DOF Robotic Exoskeletons

This project aims to develop intelligent control systems for robotic exoskeletons that can better assist human movement. Using artificial intelligence, specifically deep reinforcement learning, the project will train controllers in advanced computer simulations that model both the human body and the robotic device. Muscle activity signals (EMG) will be incorporated so that the exoskeleton can adapt its assistance based on the user’s physical effort. The trained controllers will then be tested and refined on a real multi-joint exoskeleton. The project will benefit the participating institutions by strengthening their collaboration in AI and assistive robotics, and enhancing expertise in adaptive robotic systems for healthcare and rehabilitation applications.

APEGA Science Olympics Scoring System

The APEGA Science Olympics is an annual competition that relies on a large volunteer base to evaluate student teams. Currently, the events use a paper-based judging process, distributing clipboards and scoring sheets to more than 100 judges to assess approximately 300 teams. After the event, all paper forms must be collected and manually entered and processed, resulting in significant administrative overhead and delays in producing results and providing feedback to students.
The manual review process introduces several challenges beyond inefficiency. Judges have limited ability to review or correct scores once submitted, making error correction difficult and time consuming, at the limited points where it’s available at all. Additionally, last-minute changes (such as a volunteer no-showing on the day) require manual reassignment of judges to teams, creating confusion and further increasing administrative burden. These issues all together lead to delayed feedback, increased risk of data errors (including missed teams), and overall, a process that does not scale well as the competition grows.

Analysis and experimental demonstration of fast optical switching

The rapid growth of artificial intelligence workloads, cloud computing, and resource-disaggregated services is significantly increasing traffic inside data centers, pushing conventional electrical switching architectures toward limits in bandwidth density, power consumption, and I/O scalability [1]. Electrical switching fabrics require multiple tiers and repeated opticalelectrical-optical conversions, which increase latency, energy consumption, and operational complexity as network scale grows. Optical switching has therefore emerged as a promising approach to improve scalability by routing traffic directly in the optical domain while remaining largely independent of modulation format and data rate [1-4].
The partner organization develops and evaluates high-speed interconnect technologies and network architectures for data-center environments. Its main activities include integration of optical subsystems, evaluation of emerging interconnect technologies, and identification of practical deployment paths that improve performance without increasing infrastructure complexity or power consumption. A key challenge faced by the partner is determining whether fast optical switching techniques can be implemented using commercially viable components while maintaining error-free performance and practical link budgets compatible with existing transceiver ecosystems. Broader benefits include reduced data-center energy consumption and improved efficiency of digital infrastructure supporting cloud and AI services.
This project proposes the analysis and experimental demonstration of a nanosecond-scale wavelength-routed optical switching node using passive wavelength multiplexing and routing elements such as arrayed waveguide gratings (AWGs) [1]. The work will investigate whether fast optical gating or amplification is required to support high-speed operation and how such elements affect performance and power efficiency [5]. A proof-of-concept 4×4 demonstration will be used to validate switching behavior and transmission performance at PAM4 100 GBd. The anticipated outcomes include validated
performance data, architecture guidelines for scaling toward larger systems.

Automatisation multi-niveaux par l’IA

À L’Original, notre promesse est simple : faire entrer l’art dans la vie de tout le monde à Montréal. Vous n’êtes pas artistes? Aucun problème. Notre projet Automatisation multiniveaux met la technologie au service de l’humain pour que chacune et chacun découvre, crée et s’approprie l’art en galerie, en ligne et à la maison. L’Original existe pour démocratiser l’art. La technologie est notre véhicule pour y arriver, avec une approche plus personnelle, plus fluide et plus inclusive.

RAG signifie Recherche Augmentée par Génération. Imaginez une bibliothécaire ultra rapide qui connaît nos œuvres et vos goûts, associée à une plume claire qui sait formuler des recommandations. Ensemble, elles vont chercher l’information dans nos collections, la vérifient, la résument et vous proposent des actions concrètes : quelles œuvres voir, comment les essayer virtuellement chez vous, comment rencontrer une artiste ou comment obtenir un devis pour une murale.

Notre automatisation multiniveaux orchestre plusieurs modules. Un module RAG pour le catalogue suggère des œuvres adaptées à votre budget et à votre style. Un module RAG pour le soutien répond vite et bien aux questions courantes. Un module RAG pour le marketing adapte nos contenus aux tendances locales. Un module RAG pour la relation avec le public prépare, après votre visite, un message chaleureux et personnalisé. Résultat : une expérience simple, humaine et sur mesure, dans nos espaces et sur nos plateformes numériques.

L’IA guide, l’humain crée. Nos stagiaires vous le disent : « Vous êtes les artistes d’aujourd’hui. »