AMResume
Back to featured work

Public repo · AI agent evaluation

Software Lifecycle Env

Deterministic benchmark environment for evaluating AI agents on testing, debugging, and maintenance workflows.

PythonFastAPIPydanticAgent EvaluationHidden Regression ChecksOpenEnvDocker

Overview

Software Lifecycle Env is a public Python benchmark environment that tests whether AI agents can inspect realistic engineering context, apply safe fixes, run validation, and avoid visible-case-only patches.

What I built

I built a benchmark environment where agents inspect an in-memory repository, patch files, run deterministic validation, and submit final answers against visible and hidden checks.

Technical focus

The technical focus is safe evaluation design: typed actions, deterministic scoring, hidden regression checks, anti-gaming controls, FastAPI wrapping, and reproducible benchmark-quality scripts.

Why it matters

This project demonstrates applied AI engineering beyond demos: designing evaluation systems that test realistic software lifecycle behavior.