ACEWORKS, a future mobility technology company, held a final presentation for its global internship program on July 31, where interns shared the results of projects focused on automating autonomous driving data and vehicle interface validation.
The presentation featured Eugene Lee, a Computer Engineering student at Carnegie Mellon University, and Taehoon Ryu, a Computer Engineering student at the Georgia Institute of Technology. Going beyond conventional research assignments, the two interns independently designed and implemented validation systems and automation pipelines that can be applied directly to ACEWORKS’ actual development environment.
Eugene Lee presented a dashboard that automatically validates the quality of BEV (Bird’s Eye View) and PCD (Point Cloud Data) labeling data used in the development of high-precision vision models. The performance of autonomous driving AI models is heavily influenced by the quality of labels in training data, but manually reviewing massive volumes of data presents practical limitations.
To address this challenge, Lee developed a system that examines uploaded data for errors across three levels: input values, individual frames, and sequences.
The system also enables users to visually identify anomalous frames. Validation results are provided in PDF, JSON, and CSV report formats, with the applied threshold values and lists of anomalous frames recorded for further review.
The dashboard validated more than 5,000 frames and over 42,000 labels in approximately 33 minutes, detecting more than 1,000 anomalous frames in the process. Another key feature is its flexible structure, which allows additional validation criteria to be implemented as functions, enabling the system to adapt to new datasets and validation requirements in the future.
Taehoon Ryu presented a “Vehicle Interface Validation Automation Pipeline” designed to automatically identify interface errors that may arise from differences in signal systems across vehicle models.
The project began with the challenge that software operating normally on one vehicle may produce errors on another, while manually revalidating every vehicle whenever source code is changed is impractical.
The system developed by Ryu detects source code changes within one minute, automatically builds the software, installs it on an actual interface board, and verifies both boot status and vehicle-specific test results. Validation results and identified causes of errors are automatically delivered through the company’s internal collaboration channel, while final approval and code merging remain under the developer’s control.
The pipeline currently supports vehicle models for the Hyundai IONIQ 5 and Genesis GV60. With a single code change, a total of 32 regression tests are automatically executed across the two vehicle models, comprehensively checking message cycles, signal values, CRC, counters, response delays, and other parameters.
During testing, intentionally inserted errors caused all 32 test items to fail, with the system automatically identifying and reporting the causes. Once the errors were removed, all tests passed successfully.
To ensure the system can continue to be used by the team after the internship, Ryu also developed 54 test catalogs, a vehicle model library, a validation execution engine, and documentation covering bench reconfiguration and system operation. Designed to accommodate additional vehicles and test scenarios over time, the pipeline has been established not as a one-time project outcome, but as a reusable validation asset for ACEWORKS.
The final presentation demonstrated how the global interns applied the artificial intelligence and software technologies they had studied at university to real-world autonomous driving development environments, directly contributing to improved operational efficiency and validation reliability at ACEWORKS.
In particular, their projects went beyond reducing repetitive tasks for developers by establishing a foundation for the systematic management of both data quality and vehicle software quality.
ACEWORKS plans to continue strengthening its global talent pipeline, providing outstanding international talent with opportunities to gain hands-on experience in the future mobility industry and deliver meaningful results.
The company will also connect technologies and ideas developed through its internship programs with internal R&D projects to further enhance its competitiveness in future mobility validation.