Harish P

Developed an LLM app to chat with product documents using Flan-T5, langchain, RAG and vector DB that helps agents answer customers' queries about the products while on call, which increased the response time of calling agents. Fine-tuned LayoutLVM3 model to extract billing and tax information from company invoices. Developed an app using the Azure invoice model to extract billing and tax information from company invoices, this saved 20% time for the Accounts team. Developed an ML model for generating calling leads daily, which increased the premium collection by 3%. Conducted extensive data analysis on non-payment customers using pandas, statistical modelling, and clustering analysis which resulted in actionable insights that improved premium collection by 2%. Developed an app that auto-generates 12 different analytical reports for 3 different clients by providing insights on customer behaviour, agents' performance, and premium collection analysis to support strategic decision-making. This helped the team save 33% of productive time daily. Created interactive dashboards using Dash & Plotly, and Bold BI to help decision-making in ongoing processes.

Developed and implemented data cleaning and preprocessing techniques to ensure data accuracy and integrity. Wrote Python scripts to scrape data, to feed ML models which helped the model better generalize. Built an ML model, that classifies the emails automatically based on the case type (requests, queries, complaints.. etc.). Developed an ML application to extract emotions out of the emails, this helped us better serve the customers and reduced the churn rate by 5%. Built an ML application that extracts the required details from 300+ PDF documents daily has saved the company from hiring an extra 15+ human resources. Created a shell script, that helped us to automatically backup & encrypt the production MySQL DB daily.