At KTH Royal Institute of Technology, I applied machine learning and statistics to large-scale biological data, developing projects end to end with biotech collaborators.
I designed scalable machine learning algorithms for spatial reconstruction of up to 100M nodes, improving accuracy to 96% and reducing time complexity from O(n²) to O(n), bringing computation time down from months to minutes.
I built link prediction and clustering algorithms that increased tolerance to noisy data from 5% to 30%. I also created reference-free data-quality metrics that replaced expensive laboratory validation with computational evaluation.
I released each method as a documented open-source Python package on GitHub so collaborators could rerun analyses, and wrote a STAR Protocols paper explaining their use. My work has appeared in four Q1 journals, including Nature Communications.
