About the job
Join Our Innovative Team!
As an AI Fusion Engineer at 42dot, you will play a crucial role in developing advanced perception systems for autonomous driving. Your primary responsibility will be to integrate various cognitive results generated from autonomous driving perception systems on an object-by-object basis. This involves estimating consistent and reliable states such as the position, speed, classification, and existence of objects surrounding the vehicle.
The fusion process goes beyond mere sensor integration; it entails combining cognitive results from the same sensor/multi-camera/multiple AI models/different representations (e.g., 2D, 3D) into a single interpretation using perception-level and AI-result-level fusion. You will be responsible for everything from algorithm design to real-time implementation in C++ and performance optimization.
Key Responsibilities
Design and implement object-level fusion and tracking logic using cognitive results from identical or diverse sensors (e.g., single/multiple cameras, camera-radar).
Develop fusion logic to integrate various cognitive results on an object basis (e.g., combining 2D/3D results, multi-camera integration, camera-radar fusion).
Manage object associations, track management, and lifecycle management.
Ensure consistent state estimations regarding object position, speed, classification, and presence.
Uncertainty-aware Fusion
Design fusion strategies that consider uncertainty and confidence based on sensor and AI model characteristics.
Resolve inconsistencies between cognitive results with different representations and observation points.
Enhance fusion logic to reduce duplicate objects and false positives/negatives.
Real-time Implementation and Performance Optimization
Implement real-time fusion/tracking algorithms based on C++.
Optimize performance considering latency, CPU usage, and memory footprint.
Adjust trade-offs between accuracy and responsiveness under real-time constraints.
Performance Evaluation and Tooling
Define metrics to evaluate object fusion and tracking performance.
Conduct performance analysis based on large-scale logs and scenario-specific validations.
Establish an automated performance evaluation and regression testing environment.
