Title of the talk/workshop
Building a Multi-Agent AI System for Legacy Code Migration using Python & NVIDIA Nemotron
Abstract of the talk/workshop
In this talk, we explore how to build a real-world multi-agent AI system using Python to solve a critical industry problem — migrating legacy codebases (like COBOL, Fortran) into modern languages such as Python.
We will walk through the architecture of CodeMigrator AI, a system that uses multiple specialized AI agents (Reader, Architect, Migrator, Tester, Documenter) coordinated via LangGraph to perform end-to-end code migration. The session will cover how to design agent workflows, route tasks between large and small models efficiently, and stream real-time results using FastAPI.
Additionally, I will share insights from my current role as a fresher at Cognizant, where I am working and exploring Generative AI technologies to build AI-driven applications for clients.
Attendees will gain practical insights into building scalable AI systems, integrating LLM APIs (NVIDIA Nemotron), handling structured outputs, and deploying full-stack AI applications using Python and modern web technologies.
Category of the talk/workshop
GEN AI, Machine Learning, and AI
Duration (including Q&A)
15 minutes talk
Level of Audience
Intermediate
Speaker Bio
- Speaker Bio (Brief): Ritish Kurma is a fresher at Cognizant with a strong interest in Generative AI and full-stack AI development. He has built multiple real-world AI projects including NLP systems, stock prediction models, and multi-agent AI applications. He is actively working on AI solutions for clients and exploring modern LLM-based architectures.
- Company/College: Cognizant
- Email: kurmaritish017@gmail.com
- Years of Exp: Fresher
Prerequisites(if any)
- Basic understanding of Python
- Interest in AI/LLMs and system design
Title of the talk/workshop
Building a Multi-Agent AI System for Legacy Code Migration using Python & NVIDIA Nemotron
Abstract of the talk/workshop
In this talk, we explore how to build a real-world multi-agent AI system using Python to solve a critical industry problem — migrating legacy codebases (like COBOL, Fortran) into modern languages such as Python.
We will walk through the architecture of CodeMigrator AI, a system that uses multiple specialized AI agents (Reader, Architect, Migrator, Tester, Documenter) coordinated via LangGraph to perform end-to-end code migration. The session will cover how to design agent workflows, route tasks between large and small models efficiently, and stream real-time results using FastAPI.
Additionally, I will share insights from my current role as a fresher at Cognizant, where I am working and exploring Generative AI technologies to build AI-driven applications for clients.
Attendees will gain practical insights into building scalable AI systems, integrating LLM APIs (NVIDIA Nemotron), handling structured outputs, and deploying full-stack AI applications using Python and modern web technologies.
Category of the talk/workshop
GEN AI, Machine Learning, and AI
Duration (including Q&A)
15 minutes talk
Level of Audience
Intermediate
Speaker Bio
Prerequisites(if any)