Mitanshu Khambayte

Project: Autonomous AI Data Analyst

• Designed and implemented an autonomous AI data analyst system using LangGraph multi-agent architecture to automate dataset analysis and insight generation.
• Built specialized agents for analysis planning, dataset retrieval, Python code generation, execution, and business insight generation.
• Executed LLM-driven Python code generation for dynamic analysis of structured datasets using Pandas and SQL queries, enhancing data insights.
• Developed self-healing workflows using LangGraph loops, enabling automatic regeneration of analysis code when execution errors occur.
• Integrated natural language data querying, allowing users to ask business questions without writing SQL or Python code.
• Built a secure code execution environment for running AI-generated analysis scripts safely.
• Created REST APIs with FastAPI for seamless integration with enterprise analytics platforms, improving accessibility of data resources.
• Implemented data visualization generation and automated reporting for key business metrics.
• Optimized system performance using caching and efficient data retrieval techniques.
• Containerized the system using Docker for scalable deployment in production environments.

Project: Bodhi Agentic NL2SQSystem

• Designed and implemented an agentic NL2SQL system using LangGraph multi-agent architecture, enabling automatic conversion of natural language queries into optimized SQL statements.
• Orchestrated a multi-agent workflow (Supervisor, Column Selection Agent, SQL Generation Agent) to collaboratively process user queries and generate production-ready SQL.
• Developed advanced prompt engineering and state management logic to improve reasoning accuracy and reliability of AI agents.
• Implemented semantic search using Elasticsearch to intelligently identify relevant database columns based on user intent.
• Built a Query Validation framework using PySpark to verify SQL syntax, schema compatibility, and execution feasibility before deployment.
• Designed a testing and benchmarking pipeline to automatically evaluate generated SQL queries against expected outputs for business analytics scenarios.
• Architected the backend system with FastAPI and asynchronous APIs, enabling scalable and flexible integration with enterprise platforms.
• Engineered multi-tenant architecture with client and project isolation, facilitating secure data access and enabling scalable deployment.
• Implemented real-time communication using WebSockets to support interactive query generation and feedback loops.
• Containerized the platform using Docker and CI/CD pipelines, enabling efficient deployment and environment configuration.

Project : Medical Chatbot using RAG

• Developed an AI-powered medical chatbot using Retrieval-Augmented Generation (RAG) to deliver accurate and context-aware responses from medical knowledge sources.
• Implemented semantic search with Pinecone vector database to enhance retrieval of relevant medical documents and research content.
• Integrated large language models (LLMs) to generate grounded responses from retrieved context, reducing hallucinations.
• Designed and optimized document ingestion and embedding pipelines for medical datasets including research papers and clinical guidelines.
• Implemented secure authentication mechanisms (OAuth, SSL encryption) to ensure safe access and protect sensitive medical information.
• Built scalable AI APIs and pipelines and deployed them on cloud infrastructure to support high availability and reliable performance.
• Enhanced response relevance through context-aware prompt engineering and retrieval optimization techniques.