Denny Hsu Email & Phone Number

Düsseldorf, North Rhine-Westphalia, Germany

• Development of ML compiler framework• Model optimization to speedup the runtime by developing quantization algorithms

Wuppertal, North Rhine-Westphalia, Germany

Master thesis title - "Hardware-Aware Partially-Connected Differential Architecture Search for Embedded Systems"- Applied for AI model optimization on the specific hardware platform• Case 1: Evaluated on Open Sources Dataset and on Two Different Hardware Platforms- Resulted in 8x faster latency (from 22209.71ms to 2786.85ms) on Raspberry Pi 4 and improving 0.87% Top-1 accuracy (from 92.27% to 93.14%) compared to the baseline model (PC-DARTS) - Resulted in 7x faster execution time (from 21.74ms to 3.08ms) on Aptiv's proprietary automotive compute platform and improving 0.16% Top-1 accuracy (from 92.27% to 92.43%) compared to the baseline model (PC-DARTS)• Case 2: Adopted on Aptiv‘s Automotive Compute PlatformIntegrated ES-DARTS algorithm into Aptiv’s proprietaries resulting speed-up of 1.6x faster compared to the baseline with a comparable performance

Wuppertal, North Rhine-Westphalia, Germany

Assisted to develop next generation embedded software and infrastructure for AI focused perception systems in autonomous driving• Developed the prototype in Pytorch that can be deployed in AI accelerators• Built Python scripts for ONNX models that are deployed in edge devices• Researched in state-of-the-art AI papers in Hardware-Aware Neural Architecture Search(HW-NAS) and presented to supervisors

Baden-Württemberg, Deutschland

Worked with cross-functional teams:• Worked with cross-functional teams in different regions to support the highly automated parking function of Volkswagen’s project in a new validation method, Software-in-the-Loop (SiL), and optimized the process resulting in 10 seconds reduction Designed new simulation model:• Deployed simulated software tool from ADTF and CarMaker via breakpoints on VS Code to validate CAN Signals and functions as well as analyzed traces on ASAMMDF and CarMakerOptimized the process:• Enhanced and implemented functions for automatic adjustment of CAN-Signals in C-Code to reduce the process from 10 sec to 0 sec on each parking function compared to the manual settingContinuous testing:• Developed Unit Testing and Software Testing with Google Test for nightly build-in Jenkins CI/CD to monitor highly automated parking functions efficientlyEstablished SOP:• Documented the guidelines of SiL for cross-functional teams and freshers on Bosch DocupediaExperienced with version control system:• Used Bitbucket and other software tools to maintain software versions

Master Of Science - Ms, Electrical Engineering And Information Technology

Courses: • Basic of Robotics • Controller Design for Multivariable Systems in State Space • Digital Control of Mechatronic Systems •.

Bachelor Of Science - Bs, Energy & Refrigerating Air-Conditioning Engineering

Focus on Electromechanical Integration