POST /data/uploadUpload the data
Bring CSV, Parquet, or JSON. Kautious Time validates the shape, maps your time and target columns, detects frequency, and stores the dataset.
Own the number before the meeting starts.
The old forecasting workflow makes smart teams look uncertain.
Somewhere in the planning cycle, a forecast becomes a negotiation. One team has a spreadsheet. Another has a notebook. Someone trusts last year's model because it is already in production. Someone else has a feeling about next quarter.
That is not forecasting. It is a ritual. Kautious Time replaces the ritual with a repeatable model tournament: every forecast tested, compared, ranked, and explained before it reaches the room where decisions get made.
Your first credible forecast in minutes.
POST /forecastLaunch the forecast
Pick specific models or use auto mode to evaluate multiple candidates. Run single-model jobs, multi-model jobs, or automatic selection.
GET /forecast/jobs/{id}Show the evidence
Poll progress, inspect results, download JSON/CSV/Parquet, and use metrics to explain why the selected model earned the forecast.
Let the models fight. Ship the one that earns it.
Don't bet your planning cycle on a single model. Use cross-validation and evaluation metrics to decide what works on your data, not what looked best in a vendor benchmark. When anyone asks "why this forecast?", you have the leaderboard.
- AutoARIMA
- ETS
- MSTL
- LightGBM
- XGBoost
- NHITS
- TFT
- Chronos
- TimesFM
- Moirai
- TimeGPT
- Hierarchical reconcilers
Statistical models for strong classical baselines.
ML models with lag features for tabular forecasting patterns.
Neural and foundation models for richer sequence behavior.
Hierarchical reconciliation when bottom-level and aggregate forecasts need to agree.
Comparisons scored with MAE, RMSE, MAPE, SMAPE, WMAPE, MASE, bias, CRPS, coverage, Winkler score, and pinball loss.
Built for the person who has to own the forecast.
You are not just buying a forecasting API. You are buying a more defensible planning conversation.
A better answer than “this is the model we use.”
Show which models were tested and why one won.
A cleaner review with leadership.
Bring charts, ranked metrics, and confidence intervals instead of a fragile spreadsheet.
A faster path from data to decision.
Launch async jobs without waiting on local environments or GPU setup.
A system your team can repeat.
Upload, forecast, compare, download, and audit through one API and admin dashboard.
Serious forecasting without the infrastructure drag.
Model coverage
30+ registered models across statistical, ML, neural, foundation, and hierarchical families.
Backtesting and metrics
Cross-validation and 12 evaluation metrics help separate signal from model preference.
GPU-backed execution
Neural and local foundation models run through serverless worker infrastructure instead of your laptop.
Async jobs
Long-running forecasts run as background jobs with status, progress, cancellation, and result download.
Data validation
Column mapping, timestamp parsing, numeric target checks, duplicate detection, and automatic frequency detection.
Secure access
API keys, session auth, JWT, and tenant-scoped datasets and jobs when auth is configured.
Early access is gated because setup matters
Backtested or it didn't happen.
For teams that already feel the cost of uncertain forecasts: demand planning, finance, capacity, inventory, and revenue planning. We onboard fewer teams, well.