At AirDawg Labs, I authored 28 high-difficulty Terminal-Bench 2.0 tasks covering terminal-native debugging, ML pipelines, and systems-engineering workflows. I also established isolated Docker evaluation sandboxes with hidden pytest verifiers and reference solutions for reproducible grading.
I strengthened those tasks with adversarial edge cases, Oracle/NOP baseline checks, and difficulty calibration to catch false-positive passes before release.
At Veertrip, I contributed to 10+ RESTful APIs for booking and user-management flows, adding token-based authentication and request validation. I also improved reservation database operations and integrated payment and notification providers.
I built AutoNexa-AI, an AI-powered automobile marketplace with Gemini-based image search, vehicle listings, and test-drive scheduling. I also built AuthShield, an identity and access management system with JWT authentication, bcrypt password hashing, and role-based access control.
