Usman Farid

I’m an AI Evaluation Engineer and Python backend developer focused on building evaluation systems that are correct, repeatable, and fast. I specialize in designing deterministic benchmark tasks, strict validation logic, and end-to-end AI evaluation pipelines that teams can trust.

At Turing (remote), I’ve ranked among top contributors in my POD for evaluation accuracy, correctness, and turnaround time. I contributed to Terminal-Bench by designing deterministic benchmark tasks and evaluator logic for terminal and system-level reasoning and execution, and I served as a Peer Reviewer to ensure evaluator reliability and task determinism.

I also worked on OSWorld, a GUI-focused AI evaluation platform, where I developed structured JSON-based task specifications for GUI workflows, expected states, and validation criteria. I built and reviewed Python evaluators that automatically verify GUI interactions and task completion, and I joined a special research effort to analyze Average Handling Time (AHT) while optimizing evaluation complexity and workflow efficiency.

Earlier, as a Software Engineer (Python / Rust), I built scalable RESTful APIs with FastAPI and Flask, implemented secure authentication with role-based access control (RBAC), and delivered backend systems for payment processing, analytics dashboards, and real-time tracking. I also migrated performance-critical Python modules to Rust, improving execution speed by up to 40%, while partnering with frontend and DevOps teams to streamline deployments and reliability.