Madhur Banger

Built reproducible Docker environments across 15 Python repositories, resolving complex dependency issues and achieving 90–95% test coverage through systematic Python testing pipelines.
Owned end-to-end CI/CD and validation workflows across 25–30 commits/repo, maintaining a high quality bar with rigorous code reviews in a fast-paced, distributed environment.
Contributed to Grok-4 achieving 75% reasoning accuracy on SWE-Bench — large-scale Python code evaluation across thousands of test cases, gathering and analyzing data to support model decision-making.

Mercor