Varsha Saini

- Designed and deployed a RAG-powered HSN Classification engine using GPT-4 and vector databases, reducing manual tax code identification by 80%. - Built a multi-agent conversational AI system using Gemini 1.5 Pro, LangGraph and Pedantic to enable Text-to-SQL translation, integrating LangChain agents and dynamic tool/function calling. - Implemented memory modules and semantic retrieval pipelines for sustained multi-turn interactions. - Deployed end-to-end AI workflows via FastAPI on GCP, collaborating with

• Development of a predictive ML system (Ensembling methods like Xgboost, AWS Sagemaker, Docker) following an Agile process. • Worked on creating a standard MLOps framework for deploying different ML models on AWS including data and model pipelines, drift calculation and unit testing. • Worked closely with cross-functional teams, including business stakeholders, data engineers, and domain experts, to ensure successful implementation and adoption of data solutions. • Experience in working on a CPG project fo

• HR Chatbot: NLP Based Conversational Bot for Employees Queries using Deep Learning with end to end Integration Pipeline and Deploying to Production . • Review Analysis: Play Store Reviews Analysis of Organization and its Competitors using Topic Modelling and other NLP Concepts. • Survival Analysis Project: ML project to Predict Lifetime of a Property to Rent Out / Sold Out. • Data Allocator: Built Full-Fledged Django Rest Framework Based Web Application where all the Processes were Automated. • Automati