Title of the Talk:

Autonomous AI Agents in 2026: From Execution Pipelines to Self-Improving Systems

Abstract:

By 2026, AI agents have evolved from experimental prototypes into production systems capable of planning, reasoning, and multi-step execution. However, real-world deployments reveal a consistent limitation - agents are strong at analysis, but unreliable at execution.

This talk examines the architectural gap between intelligence and action in modern agentic systems. It highlights how most failures arise not from model capability, but from orchestration issues such as weak task routing, poor context transfer, and non-deterministic execution flows.

Building on emerging industry and research trends, the session introduces a practical framework for designing reliable AI agents using structured execution pipelines, multi-agent coordination, and validation-driven feedback loops. It also explores the shift toward self-improving systems, where agents learn from execution outcomes and continuously refine their behavior.

Profile:

Nilesh Kumar is a Senior Software Engineer at Inspire Brands, USA, with over 8 years of experience building scalable, high-performance systems across mobile and cloud platforms. He specializes in modern application architecture, including React Native, TypeScript, and distributed system design.

At Inspire Brands, he works on next-generation AI systems, focusing on autonomous agents, multi-agent coordination, and self-improving architectures. His work bridges the gap between AI reasoning and real-world execution, designing reliable, production-grade agentic systems.