Andreas.

Development of an automotive header-only C++ library

C++

• Improved a performance-critical C++ API through benchmark-guided refactoring, recovering lost inlining and simplifying call sites and template constraints; reduced dynamic instruction count by up to ~17×, with comparable gains across multiple compilers.
• Achieved 2× faster runtime for trivial types and 5–20% lower memory use in embedded automotive systems by implementing a template specialization for a std::optional-like type.
• Boosted runtime performance in release builds by ~10% across 10+ customer compilers by adding move semantics to fixed-capacity containers (vectors, lists, maps), benchmarked with Google Benchmark.

Python & CI

• Built a reproducible, CI-integrated micro-benchmarking system using Docker and Python, utilizing instruction counts and simulator-based cycle counts to detect performance regressions across multiple toolchains.
• Reduced maintenance load by ~5% by automating analysis of 20+ customer repositories, identifying 50 unused public APIs and driving deprecations.

Leadership

• Fostered a customer-centric team culture by prioritizing feature requests in the backlog, organizing them into epics/stories, and implementing the first requests, leading to active developer engagement with customer needs.
• Converted a frustrated customer into a satisfied one by establishing and carrying out the Developer Relations Engineer role, directly addressing customer developers’ issues.

Task force support: Refactored GPS-Based Route Prediction Component

C++

• Reduced code complexity by 25% and eliminated 1,500+ lines in real-time radar-based situation interpretation algorithms through template-based redesign and restructuring of “iceberg classes.”

Python

• Reduced development time by 33% during a tight project deadline by implementing a Python-based automated toolchain that tested thousands of output signals for functional correctness during code refactoring.
• Accelerated rare-event detection by ~40× by implementing Python scripts to retrieve, convert, and analyze ~1,000 recorded sequences from radar-based driver assistance systems.

Leadership

• Delivered introductory technical workshops on Python and GoogleTest, supporting onboarding and enabling junior engineers to contribute more effectively.

• Developed and maintained C++ modules for camera image and metadata processing to support labeling and training workflows.

• Administered an Azure-hosted SQL database for image metadata and label storage, ensuring reliable data access and schema evolution.

• Integrated Azure DevOps and Azure ML pipelines to automate label data deliveries and schema updates.

• Expanded Azure ML pipelines and SQL-based workflows to improve data handling, scalability, and pipeline functionality.

• Developed a deep-learning–based prototype for LiDAR blockage detection in Python, achieving ~90% detection probability in staged pollution scenarios.

• Performed in-depth analysis of LiDAR sensor performance under diverse contamination scenarios, identifying key factors affecting accuracy and reliability.

• Engineered and extended a Python preprocessing framework for point cloud data, supporting multiple LiDAR types (multi-return, intensity, and other features) to streamline training data generation and augmentation.

• Created accurate LiDAR sensor models to enable comprehensive sensor-set evaluations.

Leadership

• Coordinated the development of the first multi-sensor 3D box labeling toolchain, streamlining annotation workflows across multiple teams.
• Conducted technical coding interviews and participated in behavioral/role-fit interviews for software-engineering candidates.

Projects

• Analyzed large-scale mobility data using Python, extracting insights for customer research projects.
• Implemented new APIs in the customer’s C# software stack, enabling smoother system integration.
• Debugged the Fresnel integral calculation in customer C++ software to ensure accurate road curvature computation.
• Evaluated real-time performance of charting and plotting libraries (C++/C#) on customer hardware, guiding technology choices.
• Developed C# desktop software to analyze electronic control units (ECUs).

Leadership

• Led and coordinated a team of five developers to refactor a customer’s Java software system, improving maintainability and project delivery.

• Conducted large-scale numerical computation, Monte Carlo simulation, and scientific programming in Fortran for theoretical high-energy physics, using Mathematica and other computer algebra systems for symbolic and semi-analytic manipulation of scattering amplitudes before numerical implementation.
• Published in peer-reviewed journals and presented at international conferences.
• Supervised doctoral students and contributed to collaborative research projects.
• List of publications