Evaluation Framework Development: Design, build, and maintain the foundational ADAS performance evaluation framework; e.g., data preprocessing modules, task scheduling engines, and report generation components.
Performance Evaluator development: Develop scenario-specific Evaluator modules for L2, Active Safety, NGA sub-scenarios; Develop Rule-based Evaluator as baseline, iterate with data-driven parameter tuning using collected field data, build reusable data-driven tooling, and ultimately deliver a Module-based comprehensive Evaluator;
Develop agents/skills that restructure Evaluation development workflows and collaboration modes, boosting both productivity and communication.Conduct Evaluation: using Golden Sample management and replay pipelines for both open-loop and closed-loop scenarios, covering data selection, version comparison, and regression validation.
Reporting & Frontend Infrastructure: Develop task-triggering and result-visualization frontend systems for complex evaluation and analysis; support automated generation of major release gate reports and minor version evaluation reports, plus day-to-day operations;
AI/Agent Toolchain: Leverage AI, Skills, and Agent technologies to transform development workflows and tooling (e.g., Scrum Master automation), enabling cross-functional teams to onboard quickly into our development ecosystem;
Enablement for Non-Algorithm Teams: Based on AD stack code, develop Agent-powered tools for System Engineers and Validation teams (e.g., rapid debug tools driven by test data); provide SWE56 testing tools and bench support;
