Agent Evaluation and Self-Improvement Engineer
Core
Designing evaluation strategies and self-improvement loops for autonomous agents across NVIDIA's products and platforms.
Role type
Principal Engineer (AI Agent Evaluation & Self-Improvement)
Builds
Automated evaluation systems, self-improvement loops for agents, and realistic enterprise sandbox environments.
Domain
Artificial Intelligence, Autonomous Agents, Large-Scale Systems
Deliverable
production ML models
Required skills
Experiment design (statistical significance, A/B testing), Python/Go/Rust, Large-scale system architecture, Agent evaluation frameworks, Data pipeline engineering, Technical direction
Preferred skills
Self-improving agent systems, Agent benchmarks, Reward hacking detection, Domain-specific evaluations (chip design/hardware)
Responsibilities
Build cross-harness automated evaluation checks, mine production traces for eval cases, evaluate eval datasets, create sandboxed enterprise environments, measure multi-agent swarms, calibrate model judges against human labels.
Seniority
Principal, hands-on IC
