Professional Experience

Text-to-SQL Evaluation Framework

Engineered a text-to-SQL validation framework using Streamlit and AWS Bedrock, implementing dynamic test case generation and automated execution pipelines. Integrated foundation models through Bedrock API for automated query validation, optimizing test coverage across complex database schemas.

Full-Stack AI-Powered Dashboard

Architected a full-stack analytics dashboard leveraging Next.js, FastAPI, and Snowflake's Snowpark Python API for data processing. Engineered automated insight generation pipelines integrating foundation models with Snowflake stored procedures, optimizing data analysis workflows and enabling real-time recommendation generation.

• Evaluate suitability of large language models (LLMs) like GPT-3.5-turbo, Llama2, and Mixtral 8x7B for our Consultative AI for Digital Sales response effectiveness and cost efficiency.
• Implement Prompt Engineering and Conversation Context Management strategies to optimize prompts and enhance the consultative and conversational user experience.
• Designed and deployed production-ready APIs using FastAPI for LLM behavior management, integrating seamlessly with backend services using LangChain, LlamaIndex, and ChromaDB.
• Developed a Named Entity Recognition module using PyTorch and Transformers for a conversational AI system (accuracy: 92%) fine-tuned on a custom manually-annotated dataset.

• Built a PyTorch based NLP module for automated Call Center Analytics leveraging pre-trained models with fine-tuning.
• Microservice architecture with FastAPI and Docker for NLP demo modules enhanced scalability and reliability.
• Productionized Deep Learning Models with 30% cost reduction on AWS servers.

• Designed and implemented new features in a data analytics platform.
• Integrated Python code support within Scala codebase using PySpark.
• Generated interpretation charts using Matplotlib and Seaborn.