Working on optimization and algorithms to treat depression :
• Overhauled a large-scale fMRI processing pipeline by migrating it from MATLAB to Python, reducing end-to-end runtime by >60%
• Re-engineered core image processing and neuroimaging algorithms, applying high-performance computing principles achieving 150x speedups
• Developed a new U-Net deep learning model for anatomical landmark prediction, improving average accuracy by over 70%
• Reduced runtime and cloud computing costs by ~50% by profiling and optimizing resource allocation within cloud Kubernetes cluster