Python · ETL · Automation · GIS · DACH Market
I extract, validate and automate data workflows — and I do it in German.
22 years in operations before I wrote my first line of code.
I build for the person who opens the file on Monday morning, not for the demo.
🚀 Hire me on Upwork · 🌐 Services & Pricing · 𝕏 @MR73BIIO_Q · 💼 LinkedIn · 📈 TradingView
„Każda idea, która rodzi się uczciwie w biciu serca, może przekroczyć każdą granicę, a jej czas ma większą wagę niż każda waluta."
Every idea born honestly in the heartbeat can cross any border, and its time weighs more than any currency.
MR73BIIO · Pulsus Cordis Tui · 1998
Drone imagery → GIS-ready inventory. 10k+ trees mapped from a single flight.
Python · QGIS · GeoJSON · Case study →
German invoice PDFs → validated CSV/Excel. Raw exchange feeds → gap-free archives.
Python · Pandas · Linux · VPS · Case study →
OOS split + Bonferroni correction. I ran 256 variants of my own strategy through it: 0 passed. 0 deployed.
Python · SciPy · Statistics · See the audit →
- 149 live trades audited through
audit.py→ REJECTED (saved me from deploying a false edge) - 256 strategy variants tested out-of-sample against four symbols, four market regimes, on Hyperliquid perpetuals → 0 survived the Bonferroni-corrected threshold
- 10,000+ trees mapped from single drone flight to GIS-ready cadastre
- 100% coverage on raw exchange feed → validated, gap-free time-series archive
„I set up scaffolding, not dependency: you get the code, the access and the documentation, and I step off the structure."
I do not build strategies to confirm them. I build validation pipelines to destroy them before they destroy capital.
Every candidate must pass four gates:
| Gate | Threshold | Why |
|---|---|---|
| Out-of-sample split | ≥ 100 trades | Below this, t-stat is noise |
| Edge after fees | > 0 | Round-trip cost: 0.0258% (maker+maker) |
| Statistical significance | t ≥ 2.0 on the test set | Not on the set you optimized against |
| Neighbourhood stability | ≥ 60% | Real edge is a plateau, not a spike |
On top: Bonferroni correction for 64 parameter combinations — because argmax on a grid produces false winners even on pure noise.
Result on my own strategy: 256 variants. Zero passed. Zero deployed.
Read the full audit →
- You send the mess — PDFs, raw feeds, CSV chaos, a strategy idea
- I audit & scope — 24h turnaround on feasibility + fixed quote
- I build & document — you get code, access, README, no lock-in
- I step off — the pipeline runs without me
Typical engagement: 3–5 days for ETL, 1–2 days for strategy audit.
Languages: Polish (native) · German (fluent, Swiss workplace) · English · Russian
Stack: Python · Pandas · SciPy · QGIS · Linux · VPS / systemd · Git · SQL
Credentials: Harvard CS50P (2026) · CS50 Cybersecurity (in progress)
Ready to automate the repetitive part?
Start on Upwork → ·
Browse services → ·
DM on X → ·
LinkedIn → ·
TradingView →
Pulsus Cordis Tui — time is the only currency that does not come back.