Satish Chavan

Implemented a Universal Intent Model and Sentiment Analysis using LLMs to enhance customer service analytics for telecom call transcripts. Developed a scalable model to classify customer intents such as billing issues, network complaints, and service upgrades, improving call routing and automation. Leveraged context-aware sentiment analysis to assess customer satisfaction, enabling real-time escalation for negative interactions and optimizing agent performance. Reduced manual effort in labeling by utilizing few-shot learning, improving accuracy and efficiency in analyzing vast call data. Enabled data-driven decision-making, leading to improved customer experience and operational efficiency.

• Smart Study Search(s3) Engine - Built and Deployed a search engine to find the relevant studies based on the input queries. Researchers publish medical papers which were extracted and stored in a relational Database. The input search query is being searched on this database, meaning the studies closest to the input query are displayed on the output screen.
• Medical Study Document Similarity - Compare the Similarity between Medical Studies and protocol documents. Medical Studies done by researchers and protocol documents that are prepared for validation had to be compared to match their similarity.
• Yield Output Prediction - Data from farmers of Bangladesh, Kenya, Pakistan and some middle eastern countries' data is collected for leaf production. Using this data an algorithm to predict farmers Yield Output of Tobacco Leaves in a specified area.

• Legal Document Summarization for Underwriting and Claims: This project applies NLP to summarize lengthy legal documents and contracts to assist underwriters and claims adjusters in quickly reviewing the most critical information. By using generative summarization models, key clauses, conditions, and terms are automatically extracted and condensed, allowing for faster decision-making without sacrificing accuracy. This tool improves efficiency within the underwriting and claims processes by reducing the time spent on manual document review and ensuring that the legal nuances of policies and claims are easily accessible.
• Style Transfer using GAN (PGD)-One of the complicated tasks in medical imaging is to diagnose MRI(Magnetic Resonance Imaging). Sometimes to interpret the scan, the radiologist needs different variations of the imaging which can drastically enhance the accuracy of diagnosis by providing practitioners with a more comprehensive understanding. But to have access to different imaging is difficult and expensive. To overcome this, we developed a Generative adversarial model(modified U-Net) which can generate artificial MRI images of different contrast levels from existing MRI scans.