About

I measure the performance of systems where no central telemetry collection is possible, and then try to make them faster.

ProbeLab began in 2022 as a team at Protocol Labs. In 2025, Yiannis Psaras, Mikel Cortes-Goicoechea and I founded it as its own company. We watch how global peer-to-peer networks behave in practice: how peers find each other, how long lookups take, which nodes stay online and which vanish after an hour. Then we publish it, and argue for and implement the protocol changes that the measurements imply.

In 2026 I finished a PhD in computer science at the University of Göttingen, in Bela Gipp’s group. The thesis is about measuring and optimising peer-to-peer networks, with IPFS as the case study.

Some results, each one done with the ProbeLab team or my co-authors:

• Optimistic Provide made publishing content to the IPFS DHT ten times faster and cut its network overhead by 40%. It started in my PhD, became an INFOCOM paper, and has been on by default since Kubo 0.39.0.
• My study of the Hydra booster nodes found that they added about 10% to DHT performance. Based on that, Protocol Labs switched them off in 2023.
• The NAT traversal study for IMC 2026 looked at 4.4 million hole punching attempts from 167 countries. About 70% succeed, over TCP as often as over QUIC, which contradicts the common view that UDP works better.
• Nebula, my protocol agnostic peer-to-peer network crawler. It feeds the graphs on probelab.io and the network metrics of Gnosis.

Before that: physics at Christian-Albrechts-University in Kiel, where our student team flew a particle detector into the stratosphere on the BEXUS 19 balloon. Then backend work at freiheit.com in Hamburg on the real time event backend behind Daimler’s Mercedes me app, then OriginStamp in Switzerland, then Protocol Labs.

I write most of my code in Go, and the analysis in Python. I built much of ProbeLab’s measurement infrastructure and run it today. I still build small things for fun, and some of them are on the side projects page.

If you want to talk about network measurement, or you have data you think I would enjoy, send me an email.