My research focusses on building broadly capable models of multiple modalities that behave similar to humans, i.e. collect and learn their own data. Towards this interest, I develop and interpret unsupervised post-training algorithms with applications in science and engineering.
A Cross Modal Knowledge Distillation & Data Augmentation Recipe for Improving Transcriptomics Representations through Morphological Features
I Bendidi, Y Mesbahi, A Denton, K Suri, K Dean, A Genovesio, E Noutahi ICML 2025 [paper][reviews]
How Molecules Impact Cells: Unlocking Contrastive PhenoMolecular Retrieval
P Fradkinα, P Azadiα, K Suri, F Wenkel, A Bashashati, M Sypetkowski, D Beaini NeurIPS 2024
FM4S @ NeurIPS 2024 (oral) (Best Paper Award) [paper][reviews]
On the Scalability of GNNs for Molecular Graphs
M Sypetkowski, F Wenkel, F Poursafei, N Dickson, K Suri, P Fradkin, D Beaini NeurIPS 2024 [paper][code][blog] [reviews]
Surprise Minimizing Multi-Agent Learning with Energy-based Models K Suri, X Q Shi, K Plataniotis, Y Lawryshyn NeurIPS 2022 [paper][webpage][code][talk][reviews]
Continuous Sign Language Recognition from Wearable IMUs using CapsNet and Game Theory K Suri, R Gupta CEE, Elsevier, Vol. 78, 2019. [paper][code][demo][reviews]
Transfer Learning for sEMG-based Hand Gesture Classification using Master-Slave Nets K Suri, R Gupta IEEE IC3I 2018. [paper]
Theses
Deep Hierarchical Reinforcement Learning MASc thesis, 2021 University of Toronto [link]
Application of Deep Learning and Game Theory for Sign Language Recognition using Wearable Sensors BTech thesis, 2019 Amity University [link][demo]
