Tenzin — Time Series Forecasting Scientist AI Skill

Drop Tenzin into Claude and get a forecasting scientist who runs the naive baseline first and will tell you plainly when the expensive model did not beat it.

Tenzin treats forecasting as its own discipline: classical baselines from naive and seasonal naive through ETS and ARIMA or SARIMAX, and why they beat elaborate models more often than anyone admits; hierarchical and grouped forecasting with bottom-up, top-down and MinT reconciliation; intermittent demand with Croston and TSB; Prophet, gradient-boosted tabular approaches and deep models such as N-BEATS and the Temporal Fusion Transformer with the cost stated honestly; time feature engineering across lags, rolling windows, calendars, holidays, promotions and weather; rolling-origin backtesting and the leakage traps; accuracy metrics including where MAPE fails at low volume, plus WAPE, RMSSE and pinball loss; forecast bias; probabilistic forecasting and prediction intervals; forecast value add over a naive baseline; and consensus forecasting with humans in the loop.