Welcome to v2026

This is the first note of the v2026 portfolio — a placeholder more than a manifesto.

What's coming

I'm planning to publish short technical notes on:

• Local-first AI — what I learned building Elena (Whisper + Claude on macOS) and 3DBox (Tauri + Ollama + build123d).
• Applied ML — credit scoring, model selection, and the small nuances that matter when shipping a sklearn pipeline to production.
• Product engineering — the boring parts: data quality, idempotent scripts, integrations that don't break at 3am.

# Example: a tiny preprocessing helper from the Home Credit pipeline
import pandas as pd
 
def fill_numeric_with_median(df: pd.DataFrame) -> pd.DataFrame:
    medians = df.median(numeric_only=True)
    return df.fillna(medians)

If anything here is useful to you, feel free to reach out.