At Pramana(Evident Scientific), I architected and deployed a coarse-to-fine digital pathology scanning pipeline that reduced scanning time from approximately 90 minutes to 7–10 minutes per slide. It was deployed across approximately 5,000 client slides.
I developed white blood cell detection and classification models for pathology images, including a custom U-Net and a custom ResNet. The detection system achieved 0.93 precision and 0.96 recall, while the 18-class classification model achieved 92% accuracy.
I also developed an end-to-end 3D reconstruction pipeline for serial-section histopathology images, using feature matching to align tissue structures and generating 3D meshes with contour-based interpolation and Marching Cubes.
My other imaging work included synthetic-white calibration, banding detection, tissue-fold segmentation, and out-of-focus detection. I co-authored USCAP 2026 research posters on glomerular 3D reconstruction and multi-magnification scanning of hematopoietic cells.
