The Fifth Elephant OSAI meet-up - Hyderabad edition
The Fifth Elephant
The Fifth Elephant

The Fifth Elephant OSAI meet-up - Hyderabad edition

Call for Proposals - make a submission; give visibility to your work

All submissions

Powered by VideoKen

Sonu Kumar

Modernizing Legacy Code with Multi-Agent Systems: Challenges & Lessons

Submitted Sep 30, 2025

Type of session: Hands-on workshop

When I was working at Capgemini’s innovation team years ago, one of the biggest projects we tackled was migrating COBOL systems to Java. Back then, GPT-3.5 had just come out, and like many, we were experimenting with what LLMs could (and couldn’t) do. Fast forward to today, the landscape looks completely different. With open source, domain-specific code LLMs and multi-agent AI systems, we’re no longer just asking a model to translate code—we’re orchestrating specialized agents that can analyze, review, test, and document, while keeping security checks, policy enforcement, and vulnerability scanning tightly integrated into the workflow. In this talk, Kalyan and I will share how that shift happened, and what it looks like to bring agentic AI into real modernization projects.

This session isn’t just theory—we’ll share real learnings from greenfield and brownfield enterprise transformations, where the hardest part wasn’t only converting COBOL, ABAP, or Fortran into modern stacks like Python, Java, or Node.js, but ensuring the entire SDLC pipeline was trustworthy. That meant integrating static code analysis, unit and regression test auto-generation, dependency vulnerability detection (SBOMs), and policy-driven CI/CD workflows into the modernization process. We’ll also talk about how we measured impact: developer productivity gains, higher test coverage, faster onboarding, and reduced security incidents—all contributing to measurable time-to-value. And we’ll be candid about the challenges, pitfalls, and surprises we faced along the way, so others can learn from our journey.

1-2 takeaways from our session:

  • Learn how multi-agent AI systems go beyond simple code translation to cover the full SDLC—static code analysis, security scanning, policy enforcement, automated test generation, and documentation.

  • Walk away with practical insights on measuring efficiency gains in modernization projects—developer productivity, test coverage, reduced vulnerabilities, and measurable time-to-value.

  • Understand how security and compliance guardrails (e.g., SBOM generation, dependency scanning, and policy-driven CI/CD workflows) can be built directly into the modernization pipeline to ensure safe, enterprise-grade AI adoption for software development.

Which audiences is your session going to beneficial for?

  • Enterprise architects exploring modernization strategies for legacy systems
  • Developers & engineers working with COBOL, ABAP, Fortran, or other legacy codebases
  • Engineering managers & SDLC leaders focused on productivity, efficiency, and governance
  • Security professionals interested in policy enforcement and vulnerability management during modernization
  • Open source contributors & AI practitioners curious about applying multi-agent systems to real-world enterprise challenges

Who We Are

Kalyan Prasad
Balancing Industry & Academia | Community Leader | Kaggle Advisor

  • AI & Data Science Leader
  • HYDPY, Chair, Co-Chair @ PyConf Hyderabad, PyCon India
  • 4x Kaggle, Global Speaker & Mentor
  • LinkedIn

Sonu Kumar
Serial AI Entrepreneur | AI Researcher | YouTuber (AI Anytime)

  • Built a HealthTech Company “Sporo Health” in the USA valued at $15M
  • Building Kodryx AI
  • YouTube and AI Community: AI Anytime (44k+ Subs., 3000+ Members in Community)
  • Patent Holder and Papers in Nature, Frontiers, et al.
  • LinkedIn | YouTube | GitHub
All submissions

Comments

{{ gettext('Login to leave a comment') }}

{{ gettext('Post a comment…') }}
{{ gettext('New comment') }}
{{ formTitle }}

{{ errorMsg }}

{{ gettext('No comments posted yet') }}

Register

Hosted by

The Fifth Elephant
The Fifth Elephant
Jumpstart better data engineering and AI futures

Supported by

Community sponsor

Meta
Meta