Charudatta Manwatkar

- Helped reduce cancer patient’s exposure to dangerous X-rays by 50% by generating two distinct chest X-ray images from a single X-ray image via a GAN model. The outputs passed the clinical quality criteria with 0.84 SSIM score. - Identified cancerous regions in CT scans of the brain using a U-Net like image segmentation model with custom layers. Achieved average dice score of 0.87

NLP: -- Extracted relevant information from noisy and unstructured documents via Named Entity Recognition (NER) with custom trained entity classes such as ‘Board of Directors’, ‘Place of Incorporation’, etc. using SpaCy library and Conditional Random Fields model (sklearn-crfsuite). -- Over 70% of the entities were recognized exactly and 20% partially, saving 6000 work-hours per year for an international investment bank client since the tool was deployed to production. COMPUTER VISION: --Identified structural damages to rural Indian roads by processing image and LiDAR data with traditional as well as deep learning algorithms (YoLo, U-Net). -- Over 90% of the damages were detected within the desired threshold. The project was featured at a World Economic Forum (WEF) conference in 2019.

· Worked on theoretical and numerical analysis of data from Quantum Optical experiments. · Helped in planning and execution of experiments. · Co-authored a research paper in an international peer-reviewed journal.