Education
- Supervised by Prof. Eric Kolaczyk.
- Thesis: Hybrid graph representation learning for molecular optical property prediction in low-data regimes
- Courses: Network Science, Probabilistic Graphical Models (UdeM), Statistical Inference, Advanced Distribution Theory, Regression, GLMs.
- GPA: 3.78 / 4.00
- Supervised by Prof. Mikhail Zamkov.
- NCAA Division I Men’s Soccer Team.
- GPA: 3.99 / 4.00
Experience
- Designing hybrid deep learning and tree-based models to improve generalization in low-data graph prediction.
- Exploring pretraining strategies for graph-level prediction using GNNs and Graph Transformers.
- Executing large-scale model training and evaluation on GPU-accelerated HPC clusters.
Tools: PyTorch, PyG, RDKit, DGL, HPC (Linux/Unix, Bash scripting, Slurm)
Returning full-time June 2026
- Collaborated with traders to develop a forecasting framework to predict client-level trading volume using market, macroeconomic, and seasonal signals.
- Applied regularized and tree-based regression methods, PCA, and unsupervised clustering for feature extraction and behavioral segmentation in time-series data.
- Delivered Python and KDB/q analytics pipelines integrated with the trading desk’s workflow to support daily client monitoring and strategy decisions.
Tools: Python, KDB/q, Pandas, NumPy, Scikit-learn, Statsmodels, XGBoost, Plotly
- Developed a multimodal ML model integrating image, climate-sensor, and harvest data for weekly crop-yield prediction, improving accuracy by 23%.
- Fine-tuned and deployed YOLOv5 models for seedling detection and germination tracking through an internal Flask–SQL web application.
- Automated crop-monitoring pipelines with custom CV models, expanding sampling coverage fourfold.
Tools: PyTorch, YOLOv5, OpenCV, Flask, SQL, Azure, Git
- Studied shape-controlled synthesis and photophysical properties of colloidal quantum dots for solar energy applications.
- Contributed to peer-reviewed publications and presented findings at the 2023 ACS Conference.
Tools: COMSOL, OriginPro, Excel
Funding and Awards
- Valued at $27,000/year. Awarded nationally for research potential in the natural sciences.
- Valued at $20,000/year. Awarded for excellence in research in mathematics and natural sciences in Quebec.
- Highest departmental GPA (Physics & Astronomy, Mathematics).
Projects
- Interactive frontend dashboard for personal use visualizing U.S. Treasury yield curves, spreads, and PCA-based regime analysis from official Treasury data.
- Implementation and analysis of alternative positional encoding methods in the learnable structural and positional encoding framework.
- An extension of the EdgeBank benchmark for temporal link prediction using a frequency-based sampling approach.
- Submission for the 2025 NFL Big Data Bowl.
- An implementation of Tipping and Bishop’s Probabilistic PCA and Mixture of PPCA with applications.
- An implementation of Breiman’s Random Forest and several variations from scratch.
- A Python package implementing regression estimators and hypothesis testing from scratch.
Publications
Additional Manuscript Contributions
Dulanjan Harankahage, Divesh Nazar, Korneel Molkens, Mykhailo V. Bondarchuk, Christopher M. Hicks, Andrew A. Marder, Michael Montemurri, Adam Roach, Ivo Tanghe, Liangfeng Sun, Richard D. Schaller, Benjamin T. Diroll, Anton V. Malko, Alexander N. Tarnovsky, Dries Van Thourhout, Zeger Hens, Pieter Geiregat, Mikhail Zamkov
Dulanjan Harankahage, James Cassidy, Jacob Beavon, Jiamin Huang, Niamh Brown, David B. Berkinsky, Andrew Marder, Barbra Kayira, Michael Montemurri, Pavel Anzenbacher, et al.
Jacob Beavon, Jiamin Huang, Dulanjan Harankahage, Michael Montemurri, James Cassidy, Mikhail Zamkov
James Cassidy, Dulanjan Harankahage, Jack Ojile, Dmitry Porotnikov, Lexie Walker, Michael Montemurri, Bianca S.L. Narvaez, Dmitriy Khon, Malcolm D.E. Forbes, Mikhail Zamkov