Utkarsh Jaiswal

Architected a stateful, multi-agent system utilizing LangGraph and LangChain to fully automate the end-to-end construction finance loan lifecycle.

Ensured scalable and asynchronous task processing through strategic implementation of Celery and RabbitMQ.

Engineered a custom Model Context Protocol (MCP) Server within Python SDK, designed to manage and persist context for stateful, multi-agent AI workflows.

Enabled complex financial process automation by developing robust and efficient AI solutions.

Architected and deployed a real-time election analytics dashboard using Python and Django RESTful APIs, applying Object-Oriented Programming (OOP) and Low-Level Design (LLD) principles.

Processed voter data for 175 seats, enabling data-driven decision-making for political strategies through modularization, abstraction, and separation of concerns.

Developed an AI-powered database query application leveraging Retrieval-Augmented Generation (RAG) with Gemini LLM and LlamaIndex to translate natural language inputs into SQL queries.

Integrated VectorStoreIndex with semantic embeddings and similarity search, achieving 40% faster metadata retrieval from PostgreSQL and MySQL databases.

Led the backend development team, designing and implementing a user management system from scratch using Node.js, TypeScript, and MongoDB.

Integrated Google OAuth for secure login and Twilio for OTP verification, enhancing user authentication and security.

Deployed the application on AWS EC2, achieving 99.9% uptime with a robust CI/CD pipeline built using Docker and Jenkins, which reduced deployment time by 30%.

Developed a distributed system supporting a real-time debate module with WebRTC and Socket.io, ensuring seamless multi-user discussions and low-latency communication for 100+ participants.

Engineered an interactive Angular-based module with a robust Java and MongoDB backend, enabling users to solve 5,000+ coding problems seamlessly.

Increased user engagement by 30% and ensured high performance and scalability through efficient dataset management.

Built a cloud-based code evaluation system using Judge0 compiler on AWS EC2, accepting submissions in multiple programming languages.

Provided real-time feedback on test case passing, errors, and Time Limit Exceeded (TLE) statuses for submitted code.