• Developed the backend architecture for EXOS Email Automation, creating Python APIs and workflow endpoints.

• Migrated legacy reporting to Microsoft Fabric, optimizing 43 SQL-based ETL scripts into PySpark pipelines.

• Built a scalable API automation framework for ServiceLink’s EXOS mortgage-loan platform, reducing validation time by 20-30%.

Curated assignments and provided mentorship to students from non-technical backgrounds, assisting them in transitioning into the field of data analytics through the course titled 'Intro to Data Analytics

• Spearheaded data annotation and augmentation strategies, enhancing blood vessel detection accuracy by 5% in kidney images.

• Trained advanced models (U-Net, Res-UNet) on a dataset of 300 kidney images, achieving a Kaggle baseline score with Dice and Jaccard Index scores of 0.86 and 0.87.

• Developed a dashboard using Weights and Biases, reducing training time by 30% through optimized model parameter tuning.

• Developed a web scraper utilizing the Wikipedia API and Python, extracting 400,000 sentences and reducing data collection time by 40%.

• Collaborated with Purdue University’s lab to enhance BERT word embeddings extraction, achieving a 20% reduction in output time.

• Automated the game Codenames by improving word matching efficiency, obtaining an average cosine similarity score of 0.76.

• Enhanced speckle removal in 500 RISAT-1 SAR satellite images using the POAC algorithm in PyTorch, achieving a PSNR of 44 and SSIM of 0.97.

• Integrated the POAC model into a user-friendly pipeline, increasing operational efficiency by 15% for geospatial analysis.

• Collaborated with ISRO to evaluate processed images, improving validation scores by 7% to meet quality benchmarks.