Munich, Bavaria, Germany
My software at Siemens transitioned major telcos worldwide from minute- to second-based billing. At Philips, I built a personal emergency response system, still sold over a decade later and saving many lives each year. At OSRAM, I led a team which built a multi-tenant IIoT platform used in plant production for epilepsy pharmaceuticals, and for research at NASA. At Paige, I owned the full AI/ML model lifecycle for a series of clinical cancer diagnostics products - helping pathologists and patients. At Helsing, I productize computer vision pipelines for defense, with soldiers' lives depending on it. My primary language is Python (20+ years; incl. PyTorch, cuML, scikit-learn), and I also use Rust & Go. My cloud experience centers on AWS, with some Azure exposure. Most of my professional experience was at large multinational enterprises, in cross-functional teams (and often also leading them), and my core strengths are in driving systems from research to production, engineering of AI/ML/platform infrastructure, performance profiling and optimization, tenacious troubleshooting and proactive automation.
Working with research scientists on productization of a computer vision & telemetry pipeline for defense applications, with focus on system integration, profiling & optimization. - Optimized the computer vision processing pipeline to operate at 2.8x of the original framerate - Implemented a range of features: telemetry acquisition, actuation integration, evaluation ETL pipeline, safety harness & overrides - Introduced automated regression monitoring of algorithm performance metrics along with trend tracking & visualization - Raised the bar for dependency management and code structure quality; championing multifaceted code checking, automation, in-depth code reviews, proactive knowledge sharing, deep refactoring practices and caps on cognitive complexity of the code
Drove full development lifecycle of AI/ML clinical cancer diagnostics products, from research to production and through development of AI/ML tooling/infrastructure: - Transformed packaging of trained models from a manual, multi-sprint, team-wide process into a one-day, single engineer operation through implementation of templates for configuration of model inference and transformations of its outputs, which scaled teams to more products. - Automated regulatory acceptance tests (reused across teams), further reducing error-prone manual release tasks. - Promoted to a product owner (2023) and led a cross-functional clinical product team of 5 engineers & scientists (and coordinating with medical, UX & frontend, infrastructure & regulatory staff). - Launched & collaborated on many products, incl.: prostate & breast cancer biomarker detection, prostate / colon / pan-cancer detection, classification and segmentation, model-based tissue segmentation tool (with superior accuracy & distortion robustness), and contributed to the Virchow pathology foundation model. - Packaged & released all models to production under FDA/CE-IVDR/ISO certification requirements. - Troubleshooted (profiling, log analysis) performance & memory issues from training (DGX A100/V100 cluster) to production inference (AWS T4 instances); applied model compilation through TensorRT, reducing latency and cutting the operational costs by 15%. - Orchestrated experiments & model training across an on-prem cluster alongside data scientists, improved collection & visualization of metrics, built an LLM-assisted data labeling workflow, etc. - Mentored junior engineers and advised peers (e.g. code reviews, pair programming, best practices, CI/CD, idiomatic Python), eliminated majority of team's tech debt, maintained shared ML libraries. - Collaborated with major pharma partners & shipped biomarker evaluation results for their research.
Lead software engineer in the Fluence unit of OSRAM, on the Smart Farming team tasked with digitization of vertical farming and greenhouse operations. - In a Senior role, I led a team of 3 to develop a Node.js/Electron app for PHYTOFY research light control, later also implemented and open-sourced a CLI - used by e.g. NASA, 3rd-party integrators (e.g. climate chamber manufacturers), for remote control at the Customer Cultivation Support Lab, and to automate the factory system tests. - Advised the division CEO office on 3rd-party partner & systems selection in relation to merger/acquisition efforts and lab automation operations. - After being promoted to Principal, I led a team of 9 engineers and carried out dual roles of a product owner and an engineer. The team was tasked to develop a multi-tenant IIoT platform on AWS, tailored to operating indoor / greenhouse plant production, integrating 3rd-party vendors to provide lighting actuation, sensor data acquisition & visualization, and decision-support tools leveraging satellite-based climate lighting models. This covered: cloud infrastructure, backend services for user and tenant mgmt., charts/dashboards, alerts/notifications, actuation scheduling, web frontend & mobile app. - Mentored/co-developed with the broader R&D team, raising the bar for the engineering & DevOps practices enabling the move from PoCs to launching products, incl. e.g.: joint troubleshooting across the full stack, knowledge sharing w.r.t. AWS services and deployment, application of CI/CD automation, introduction of postmortems for production issues, UX unification across frontends, establishment of a joint backlog / roadmap improving team's coordination, resolution of the technical debt allowing the platform to scale, migration the time-series database to a managed service improving the cost-effectiveness, platform integration with enterprise security scanning infrastructure.
Senior member of the Testing and Reliability Engineering team tasked with R&D of the software engineering tools used by the mobile and cloud business units to accelerate their work and improve the quality of their output. - Implemented components of the Huawei-internal cloud-native IDE (e.g.: single sign-on integration) and its services for software testing and static code analysis (including optimization of services for short start-up time and low resource consumption) which simplified the on-boarding of new team members. - Introduced the team to CI/CD automation, dockerized the testing environment, automated all unit and end-to-end tests implemented to date, along with the build and deployment of the CI/CD runners. - Led a team of 4 external contractors and coordinated with a wider distributed Huawei team.
Designed and implemented new Philips products and services, initially for consumer electronics and video streaming, later for personal fitness and healthcare. Projects combined various languages, technologies, and spanned from resource-constrained IoT devices to backend/cloud platforms. - Developed techniques for profiling (code instrumentation, OS event monitoring etc.) which allowed to optimize the cache memory and memory bus bandwidth footprint of a H.264 decoder on a multicore TriMedia DSP CPU. - Implemented temporal scalability QoS techniques (for Linux kernel and userspace) for streaming (encrypted) H.264 video, used by e.g. STB810 set-top-box and VP5500 VoIP phone. - Prototyped an OpenGL transcoder for the cloud gaming Games@Large EU project. - Optimized frame buffer access in a specialized LED display (Lumalive) raising framerate by an order of magnitude (and enabling video playback). - Implemented social features (competition, cooperation) for the DirectLife fitness tracking and created prototypes for monitoring of all pillars of physical fitness. - Implemented the software (C on MSP430 and ARM Cortex M0; Python services on GCP) for the GoSafe PERS device incl. a custom fusion of connectivity and location positioning tech. Later reimplemented the functionality as an Android app in a hackathon challenge. - Designed and implemented a fall prevention system - integrating an ML model (compiled Matlab code) with an ETL pipeline (Python; custom-designed compression, DSP preprocessing, etc.), deployed on AWS, scaling in a cost-effective manner to min. 100k users and operating on data ingested from Lifeline GoSafe devices. - My responsibilities also included: performance optimization, technology evaluation and selection, IP creation and circumvention, acting as a Scrum master, leading junior software engineers, mentoring PhD students, cooperation with EU projects and standardization bodies.