Suman Kumari

PROFESSIONAL PROFILE My research focuses on applying deep learning to understand and predict complex biological phenomena. I am particularly interested in modeling high-dimensional biomedical data, such as sequencing and -omics data, physiological signals (EEG, ECG), and biophysical signals (e.g., mass spectra), to uncover mechanisms behind cell behaviour, identify biomarkers, and guide drug discovery. Ultimately, my goal is to leverage deep learning to tackle fundamental problems in biology and neuroscience. EDUCATION BS Computer Science School of Electrical Engineering and Computer Science , Islamabad , 3.74 (2026) INTERNSHIP EXPERIENCE Machine Vision and Intelligent Systems Lab 01-Jun-2025 - 22-May-2026 1. Working under Dr. Moazam Fraz and Dr. Naseer Bajwa on project ”Transforming Clinical Decision-Making: Predicting Antimicrobial Resistance from MALDI-TOF Data”. 2. Preprocessed and performed binning on more than 300,000 mass spectra. 3. Trained and evaluated multi-label species-specific 1D-CNN models on diverse species–antibiotic pairs, achieving AUPRC scores between 0.7 and 0.9 through cross-site validation and fine-tuning on external datasets, outperforming baseline methods. 4. Extended the species-specific models to a single unified multi-modal model by combining the spectra and drug embeddings. 5. Reviewed research papers and documentations to learn Probabilistic Deep Learning and Continual Learning for application in the project. 6. Extended single model to bayesian model to estimate epistemic uncertainty. 7. Manuscript for these results is under preparation. 8. Extending the current work to training under continual learning paradigms and exploring strategies like replay buffer, elastic weight consolidation to adapt to evolving bacteria without catastrophic forgetting. MITACS Globalink Research Internship | York University, Toronto, Canada 08-Jun-2026 - 31-Aug-2026 Selected for Research Project: Foundational Models for Single Cell Omics: Adapting BERT for Contextual Biological Data Analysis Supervisor: Dr. Kaiqiong Zhao China Pakistan Intelligent Systems Lab 25-Jul-2024 - 15-Jan-2026 Worked under Dr. Seemab Latif and Ms. Iram Tariq Bhatti on EEG-based Auditory Attention Detection; achieved 4th place out of 43 international teams in the ICASSP 2026 EEG Auditory Attention Decoding (AAD) Challenge. FINAL YEAR PROJECT Transforming Clinical Decision-Making: Predicting Antimicrobial Resistance from MALDI-TOF Data Developing deep learning models for antimicrobial resistance prediction from bacterial mass spectrometry data, with uncertainty quantification and continual adaptation to emerging data. TECHNICAL EXPERTISE Languages Python, C/C++, Java, Latex

My research focuses on applying deep learning to understand and predict complex biological phenomena. I am particularly interested in modeling high-dimensional biomedical data, such as sequencing and -omics data, physiological signals (EEG, ECG), and biophysical signals (e.g., mass spectra), to uncover mechanisms behind cell behaviour, identify biomarkers, and guide drug discovery. Ultimately, my goal is to leverage deep learning to tackle fundamental problems in biology and neuroscience.