AI Teammates for Software Engineering: Multi-Agent and Self-Improving Coding Systems
2. Shweta Garg, Senior Applied Scientist, AWS AI Labs
Abstract:
Coding agents are moving from single-shot assistants toward collaborative, adaptive systems that reason, coordinate, and specialize. Rather than one model solving everything, emerging architectures resemble engineering teams where agents critique, refine, validate, and negotiate solutions. In this talk, we discuss recent coding agents AWS has launched, AWS Transform for accelerating modernization and reducing technical debt, and Amazon Kiro for structuring AI-driven development from prototype to production. Together they illustrate a future where coding agents are multi-agent, context-aware, and increasingly self-improving systems that do not just generate code, but organize work, coordinate expertise, and elevate how software is built and maintained.
Biodata:
Speaker 1:
Dr Behrooz Omidvar-Tehrani is a Senior Applied Scientist and Science Manager at AWS. He earned his PhD from the University of Grenoble Alpes in 2015. He leads a team at AWS AI that develops LLM agents for code transformation. He has published 60+ papers in top venues and delivered talks at leading institutions. He has delivered 7 tutorials so far in venues such as The Web Conference, KDD, SIGMOD, and CIKM. He is also a co-organizer for AAAI 2026 workshop on next-gen coding agents.
Speaker 2:
Shweta Garg is an AI Research Lead driving applied research in code intelligence and interactive agents for developer productivity, powering products such as Kiro IDE and Q-CLI. With over 13 years of experience in AI and machine learning, her work integrates reasoning, memory, and planning for collaborative coding agents. She has published in premier venues including NeurIPS, EMNLP, and ECNLP, holds several patents, and frequently delivers invited talks at industry and research forums such as KDD, GitHub Copilot, Adobe, Uber, and the Grenoble Informatics Laboratory (LIG). Her leadership combines scientific rigor with real-world impact in building deployable AI systems.