Experiment Registry

Without a registry, you overwrite the best model every time you retrain. Content-addressed storage - where hash(config) determines the storage path - makes every experiment reproducible and comparable without manual bookkeeping.

The Problem

A quant runs 50 model configurations. Results go into model_v2_final_FINAL.pkl. Next week, a new run overwrites it. The team cannot answer: which hyperparameters produced the best IC? Was that before or after the feature change? Did we already try alpha=0.01? Without structured tracking, experiments are lost, repeated, and unverifiable.

The Pattern

WRONG

import pickle

# Overwrite on every run - no history, no comparison, no provenance
model.fit(X_train, y_train)
with open("best_model.pkl", "wb") as f:
    pickle.dump(model, f)