Spatio-Temporal Transformer for pedestrian trajectory forecasting — MOT20/MOTSynth, Social-LSTM comparison and statistical evaluation.
Home directory: /nick/about
Nick Grebe
Hi, I am Nick. I build data-driven AI systems.
Data Science and Artificial Intelligence student in Maastricht, Machine Learning Engineer at Novitas Media Consulting, and interested in Geo AI, Edge ML, statistics, and good music.
C:\PERSONAL\WEB
> booting nick_grebe.exe
> loading maastricht_bachelor.dat
> syncing novitas_ai_systems
> status: online
01 / profile
Biography
I was born in Berlin and grew up all over the place. I spent most of my childhood in Perth, Australia, before moving back to Germany for secondary school. During that time I spent a year abroad in Pécs, Hungary, where I got to know the language, the culture, and myself a lot better. In 2023 I graduated from Marie-Curie-Gymnasium Neuss with a bilingual Abitur. Today I am studying Data Science and Artificial Intelligence at Maastricht University while working part-time at Novitas Media Consulting.
-
Berlin
Born in Berlin and used to the idea that home can be more than one place.
-
Perth
Spent most of my childhood in Perth, Australia, gaining distance, perspective, and an international everyday life.
-
Pécs
Spent a year abroad in Hungary, learning the language and culture while getting to know myself better.
-
2023
Graduated from Marie-Curie-Gymnasium Neuss with a bilingual Abitur.
-
Today
Bachelor student in Data Science and Artificial Intelligence at Maastricht University and part-time Machine Learning Engineer at Novitas Media Consulting.
02 / stack
Focus & Interests
- Geo AI
- Edge ML
- Statistics
- Open-source data
- Remote sensing
- Earth observation
- Explainable AI
- Machine Learning
- Good music
03 / demos
Projects & Labs
Small demos, notes, and workbenches around statistics, machine learning, and AI systems.
GITHUB / pinned projects Current work on GitHub Code, experiments, and research – straight from my pinned projects. Pedestrian-Trajectory-STTvsLSTM · cage-ai-arena · Monte_Carlo_Cluster_LP · webscrape-wPythonAnywhere Hide projects
-
-
A playable Java AI arena for benchmarking Alpha–Beta, configurable MCTS, and ResNet policy/value agents through reproducible Elo tournaments.
-
Geometry-aware adaptive sampling of primal and dual LP candidates, scored by feasibility, duality gap, complementarity, and recurring active-set support, with the goal of discovering useful active-set structure.
Topics: linear-programming · monte-carlo
-
An easy webscrape for a hypothetical website to find websites in a city with certain districts.
Data monitor
Open
Jabuka/Pomo Pit Monitor
A cautious view of public vessel-activity data across the Adriatic fisheries restricted area's three zones.
Live demo
Open
Cool Stats
A Central Limit Theorem demo with real simulations, histograms, and clear data intuition.
Live demo
Open
Mini ML demo
A small demonstration of essential Machine Learning algorithms