Sowmipriya Rajendiran

Built RAG-powered chatbots using LangChain for intelligent document retrieval and conversational AI applications. Collaborating on training a large language model, focusing on crafting open-ended prompts and evaluating model performance and response quality. Responsible for debugging machine learning code and preparing it for deployment in batch processing pipelines. Reviewed and evaluated model outputs and supporting code using structured criteria, delivering clear technical feedback and contributing to quality improvements across LLM training workflows.

￾ Contributed to the development of inhouse AI moderation systems using state-of-the-art language models (LLM), focusing on automated message interception, restricted word detection, and blocking based on ML model scores. ￾ Developed an IAB classifier capable of accurately categorizing various web pages from scratch (crawling to scoring). Included a long-former model, to enhance classification accuracy and efficienc

￾ Created and maintained rich interactive visualizations through data interpretation and analysis integrating various reporting components from multiple data sources. ￾ Built predictive models to forecast defect and breakdown patterns by comparing vehicle data to warrant

￾ Developed realtime price changes algorithm capable of refreshing product prices every 2-5 minutes automatically. ￾ Payment and address detection using ML and RE techniques - train and test the models and responsible for better performance of models. Building large scale web crawlers to fetch and parse data from different