Pinned Loading

1. demand-forecasting-backtest demand-forecasting-backtest Public

   Honest forecast evaluation: rolling-origin backtesting, baselines that must be beaten, and quantile forecasts scored on interval calibration, not just point accuracy

   Python

2. event-metrics-reconciler event-metrics-reconciler Public

   Reconciling the same business metrics from two lossy event sources: canonical definitions, dirty-data guards, MAX-merge without double-counting, and invariant checks

   Python

3. sherloq-mock sherloq-mock Public

   Agentic SQL co-pilot for BI dashboards: checkpoint-driven human-in-the-loop workflow, retrieval-grounded query generation, read-only guardrails, row-level ACL, budget-guarded LLM calls

   Python

4. g2p-benchmark-harness g2p-benchmark-harness Public

   Grapheme-to-phoneme evaluation harness on CMUdict: WER/PER metrics, model-agreement confidence scoring, and a human-in-the-loop verification queue where targeted labeling beats random

   Python

5. probabilistic-forecast-scoring probabilistic-forecast-scoring Public

   Distributional forecasts scored with CRPS and PIT calibration, plus block-bootstrap tests that separate leaderboard skill from luck

   Python

6. name-pronunciation-ranker name-pronunciation-ranker Public

   Context-aware ranking of name pronunciations: locale priors, Bayesian updating from confirmations, person-level pinning, and confidence-gated serving. Hand-curated ARPAbet/IPA seed lexicon

   Python