Hamburg, Germany
Web developer from Hamburg. I’m doing my Master’s in Computer Science with a focus on machine learning at HAW Hamburg.
Most of my work so far has been backend and infrastructure: the services behind the page, the clusters they run on, and occasionally the firmware underneath.
I’ve spent time on backend services, graphics, WebAssembly and embedded systems. Machine learning is my current focus, and where most of my time goes.
Browser 2.5D fighting game whose arenas are Gaussian-splat scans of real places — film a room, then fight in it.
A deterministic combat core drives rollback netcode over WebRTC, and a self-hosted GPU pipeline reconstructs uploaded videos into arenas — and a few photos into a rigged, playable fighter.
Bachelor thesis: a GRU and a Transformer learn to dream Gymnasium's CarRacing-v3, and a PPO agent trains inside the dream.
A VQ-VAE turns each frame into 16 tokens, two 40M-parameter world models predict the next ones, and a Dyna loop alternates real driving with dreamed driving. The GRU won.
Cocktail mixing machine built for an IoT project at Metropolia University Helsinki, using embedded C and FreeRTOS on a Raspberry Pi Pico W.
Every part of the machine is its own FreeRTOS task, two pump tasks pour both liquids at once, an RFID card pays with credits, and a web app tracks accounts and bottle levels.
Native Android app for managing your fridge, creating shopping lists, and finding recipes.
Product tagging is handled by a custom ML model I trained and serve from the cloud.
Synchronised video-watching for friends, with real-time chat and shared playback controls.
Built on a microservices architecture and deployed to a bare-metal Kubernetes cluster with a full CI/CD pipeline.
Also
Foodsharing website built as a university group project. Nominated for the Karl H. Ditze award.
More on GitHub.
Solid — daily. Muted — shipped at least once.
Open to machine learning roles and other engineering work. Email is the best way to reach me.
contact@jesseguenzl.com