Hey, I'm Philipp!
Software engineer who loves building cool stuff, at home in cloud, data science and apps.
// 01 · iOS & Mac · Swift · CoreML · Metal
My software journey started with apps for iPhone and Mac.
I shipped my first apps right after starting computer science in 2016. Lautstärke, which measures environmental noise, quickly passed 10,000 downloads, and as a freelancer I built apps for clients like Mitteldeutscher Rundfunk (MDR). On-device machine learning was the hot new thing back then, so I put it to work in prototypes like 3Dify, a parallax shader driven by disparity maps, much like the 3D photo effect on today's iPhones. 3D and design are a passion of mine anyway: I spent half a year as a Cinema 4D artist at a media agency.
// 02 · Cloud infrastructure · KVM · Kubernetes
Today, I help build Europe's next-generation sovereign cloud at SAP.
After my PhD at TU Dresden, I joined SAP Cloud Infrastructure to work on CobaltCore, the new KVM stack taking over 20+ datacenters around the globe and more than 100,000 virtual machines. My Kubernetes components connect the OpenStack interface our customers use with IronCore, a cloud-native bare-metal operating system. I work on central pieces like the VM operator and the KVM node agent, and on Cortex, which you'll meet next.
// 03 · Scheduling · Go · Kubernetes
I'm the lead contributor of Cortex, the scheduler that decides where SAP's virtual machines run.
I built this Kubernetes operator from the ground up to make smart placement decisions for almost anything: OpenStack compute (Nova), storage (Cinder, Manila) and networking (Neutron), and even gang scheduling of GPU workloads on Kubernetes. I shaped its core: a knowledge database that ingests metrics and other data sources, and a scheduler built on Kubernetes-native inventory like VM and hypervisor CRDs. Today, Cortex gives SAP Cloud Infrastructure capabilities it never had before: automated failover partitioning and reservations, committed-resource scheduling, and capacity planning through an infrastructure dashboard.
// 04 · Agentic engineering
I'm always looking to improve the things around me. Lately, that means agents.
Trying new technologies and ideas is what I love doing, and recently I've focused on how agents can automate daily work and bridge knowledge gaps. Around CobaltCore, I've set up agentic pipelines for code review and release management, and for keeping our issue board and progress tracking current through a knowledge graph of issues and commits. Just as important is keeping them safe: a network-based sandbox guards against prompt injection and leaks of sensitive company data. I'm a coder at heart, and working with AI helps me grow and learn.
// 05 · Data science · Monitoring · Perses
I love collecting data and finding out what it can teach us.
My PhD gave me a strong background in data science: trajectories and user data, plus large amounts of geodata and real-time signals. I enjoy digging for patterns, testing hypotheses, and measuring whether an idea actually has an impact. In the CNCF project Perses, I help push open monitoring forward; my canvas panel draws graphs straight from query data, like failover partitions or network throughput. I look for where cloud-native and data science meet, and off the clock, I analyze my finances in Grafana.
// 06 · Machine learning · TU Dresden
Classical and deep machine learning problems are my jam.
Ten years ago, I trained my first recurrent network for sequence prediction, and it failed miserably on Romeo and Juliet. I kept at it, picking up theory and practice along the way, from building balanced datasets to unsupervised learning. For my diploma and my PhD at TU Dresden (Dr.-Ing., 2024), I developed models for transport mode detection on phones and for error-robust, geometry-based traffic light matching, and published them at top international conferences.
- Cloudy with a Chance of Green: Measuring the Predictability of 18,009 Traffic Lights in HamburgIEEE IV 2024
- GeoAI-Powered Lane Matching for Bike Routes in GLOSA AppsACM SIGSPATIAL 2023
- Matching Traffic Lights to Routes for Real-World Deployments of Mobile GLOSA AppsIEEE ISC2 2022
- Selecting Resource-Efficient ML Models for Transport Mode Detection on Mobile DevicesIEEE IoTaIS 2022
// 07 · Product engineering · PrioBike
Platform engineer by day, product engineer at heart.
I love working with smart people on polished solutions, whether that's an architecture, a clean package structure, an interface or an RFC. What drives me is impact: working back from what users need to the concrete implementation, end to end. In my PhD project PrioBike, I led a small team building a green wave app for cyclists in Hamburg, from the machine learning behind smart routing and traffic light matching all the way to the user interface, which won 3rd place at the Deutscher Fahrradpreis.
Cloud, data or something new? Let's build something together.
I'm based in Dresden and love collaborating with people all over the world.
Imprint Dr.-Ing. Philipp Matthes Clara-Viebig-Str. 9, 01159 Dresden hi@philippmatth.es