ArchetypeXI: Football Player Archetype and Similarity Platform
Data Engineer (solo project) · ~2 months (Sep–Oct 2026) · 2026
Processes per-90-minute stats for 1,827 players into 35 features, clusters them into 5 playing-style archetypes with K-Means, and serves them through a Streamlit app.
Pipeline flow
Source
per-90-minute stats
1,827 players
Transform
Python, scikit-learn
35 features (VIF filtering, scaling); K-Means 5 archetypes; cosine-similarity engine with pytest tests
Storage
Parquet
feature storage
Serve
Streamlit
multi-page app on Streamlit Community Cloud
TL;DR
- Per-90-minute stats for 1,827 players are processed into 35 features (VIF filtering, scaling, Parquet) and clustered into 5 playing-style archetypes with K-Means.
- A cosine-similarity engine with pytest tests, served through a multi-page Streamlit app.
Metrics
- 1,827players processed
- 35features per player
- 5playing-style archetypes (K-Means)
Stack
- Python
- scikit-learn
- UMAP
- Parquet
- Streamlit
- pytest
Problem
[situation before the project / motivation — not in the CV, fill in yourself]
Architecture & technical decisions
The feature pipeline (VIF filtering, scaling) is stored as Parquet; K-Means clustering into 5 archetypes; the cosine-similarity engine is tested with pytest; served through a multi-page Streamlit app. [add rationale + trade-offs]
Data model
[short explanation — fill in if any]
Data quality & monitoring
The similarity engine is verified with pytest tests. [complete if any]
Code snippet
# [5–20 lines of the most representative Python — pick from the repo]
[why this code is interesting]
Challenge
[the hardest problem and how you solved it — fill in yourself]
Result
Publicly deployed app at https://archetype-xi.streamlit.app.
What I learned
[1–2 lessons — fill in yourself]