At AfterQuery, I create Go, Python, and C/C++ benchmark tasks for AI coding agents, including bug fixes, memory-safety tasks, and browser-tested web-app repairs. More than 25 tasks have been approved.
I package tasks in reproducible Docker environments with reference solutions and deterministic verifiers. I calibrate difficulty by reviewing model runs and distinguishing task difficulty from environment failures or loose tests.
At AfterQuery, I also built a Node.js and TypeScript validation workbench with 179 automated tests. It runs broken and fixed versions side by side in Docker and cut task pre-validation time from 645 to 349 seconds.
Earlier, I built reusable Next.js, React, and TypeScript components as a Full Stack Development Intern at CI&T, and replaced Mengoni Engenharia & Arquitetura’s manual invoice intake with an AppSheet app and JavaScript automations.

