Hi there, I am Niklas.
I like figuring out how things work.

PhD candidate in machine learning. I spend most of my time building systems that learn from data - and occasionally go down rabbit holes trying to understand why they work at all.

Email GitHub LinkedIn Scholar Visit my Blog

About Me

My PhD is on teaching ships to navigate themselves - specifically, getting reinforcement learning agents to handle the messy reality of inland waterways, where the currents are unpredictable, the data is noisy, and running aground is not an acceptable failure mode.

At some point I also spent nine months as the founding engineer at a mobility startup, which turned out to be a crash course in everything nobody warns you about: building infrastructure from zero, making architectural decisions that have to survive contact with reality, and learning that "neatly packaged problem" is mostly a myth.

The blog is where I follow whatever happens to be interesting that week. Ships, signal processing, Fourier transforms — I tend to start with "I wonder how that works" and end up three papers deep. Fair warning.

Expertise

Deep Learning 90%

Python / Torch 85%

Data analytics 70%

Ops & Deployment 60%

Infrastructure 50%

Documentation & TeX 70%

Selected Work

INDUSTRY

Nox Mobility: Founding Engineer & Tech Lead

2025

Shaped the technical foundation as founding engineer for a mobility startup from zero to investor-ready. Designed and implemented full-stack architecture: containerized microservices, data pipelines, multi-database systems, web scraping infrastructure, CI/CD, and analytics dashboards. Followed the ride for 9 months from zero to seed, defining architecture, security, DevOps strategy, and product-market-fit.

Python OpenAPI Docker PSQL Airflow Bare Metal CI/CD System Architecture

Visit Website →

RESEARCH

2-Level Reinforcement Learning Framework

2025

A modularized DRL framework for autonomous vessel control on inland waterways, featuring separate agents for high-level path planning (considering traffic rules and dynamic obstacles) and low-level path following. Outperforms traditional APF and PID controllers by 65% in obstacle avoidance while reducing control effort.

Deep RL Path Planning COLAV Autonomous systems Big Data Real-World Validation

View Paper →

RESEARCH

AIS Trajectory Extraction Framework (α-method)

2024

An open-source Python framework for extracting clean ship trajectories from noisy AIS big data. Uses maneuverability-dependent α-quantile-based filtering to handle technical inaccuracies and compliance issues in raw AIS records. Robustly extracts long, uninterrupted trajectories for maritime domain awareness and algorithm testing.

Automatic Identification System Big Data Automatic Processing FOSS

View Paper →

RESEARCH

Bootstrapped DRL for Robust Path Following

2024

A bootstrapped Deep Q-Network (DQN) controller for autonomous vessel navigation on restricted inland waterways. Handles challenging conditions like high flow velocities and shallow banks on the Rhine. Trained in diverse, realistic environments and validated against real-world data, demonstrating superior adaptability compared to vessel-specific PID controllers.

Deep RL Path Follwing Autonomous systems Real-World Validation

View Paper →