Sneha Mohite

Contributed to the ongoing development of a Retrieval-Augmented Generation (RAG) chatbot for HR policy support by improving chatbot responses and automating the evaluation of its outputs using custom quality metrics and DeepEval Library. Built a real-time HR policy ingestion pipeline using Azure Functions and Azure AI Search, enhancing document indexing and retrieval efficiency. Built a feedback loop to continuously improve the chatbot's performance and user experience through model refinement.

Developed a Natural Language to SQL chatbot (PoC) using SQL Agent, enabling intuitive, conversational data querying from structured databases. Created a Website Search Chatbot (PoC) integrating Google Search functionality, improving content discovery through natural language queries.

Architected and led the backend development of the second iteration of a text and image search platform, improving retrieval efficiency by 10% and increasing image search accuracy by 58%.

Developed and deployed an end-to-end ML pipeline for quality control task prediction on Databricks and Airflow including system architecture design, model development, integration, and

production deployment, resulting in annual savings of $150M through optimized predictive processes and automation. Automated PDF document comparison with AI-powered models, boosting validation accuracy by 40%. Built custom object detection models with Google Document AI, achieving 85% classification

accuracy for artwork components. Designed and deployed Power BI dashboards to visualize ML model metrics and business KPIs, enabling data-driven decisions.

Fine-tuned BERT, GPT-3, and GPT-Neo models to analyze employee surveys, extracting actionable insights to inform HR strategies. Automated feedback summarization, improving engagement by providing deeper insights to

management. Architected a sentiment analysis solution using BERT, achieving granular feedback accuracy through optimized preprocessing pipelines. Built automated data preprocessing workflows with Azure ML, accelerating dataset preparation for ML models.

Developed a client deal prediction model with XGBoost, achieving 78% accuracy in forecasting outcomes for business growth. Created a PoC for invoice data extraction with Faster R-CNN and DeepSpeed, enhancing document processing efficiency.