Utsav Sharma developed an Autonomous Software Engineering Agent for the google/adk-samples repository, focusing on automated task solving within software engineering benchmarks. He implemented the agent in Python, leveraging Docker to create a reproducible, containerized benchmarking environment that supports isolated and scalable experimentation. The agent autonomously addressed tasks from SWE-bench and TerminalBench, enabling automated evaluation in CI-like workflows. Utsav’s work established a robust foundation for iterative benchmarking and experimentation in machine learning-driven software engineering. His contributions demonstrated depth in Docker orchestration, Python development, and software engineering practices, with collaborative development reflected in co-authored commits and integration into the project’s workflow.