Moirai frontend
-
Updated
Apr 4, 2026 - TypeScript
Moirai frontend
Example of different time series forecasting models and processing methods.
This project was driven by the need to provide our client Mela, a startup focusing on simplifying cryptocurrency investments, with a reliable and scalable platform for backtesting trading strategies. Additionally, leveraging advanced forecasting methods would enhance the platform's capability to predict future price trends with greater accuracy.
Self-hosted forecasting + prediction service. Five zero-shot time-series foundation models (Chronos-2, TimesFM 2.5, Moirai-2, Toto-1, Sundial) across six forecast types, plus nine supervised tabular ML backends (LightGBM, XGBoost, sklearn family) with calibrated / stacking / diversified meta-learners. Unified REST API + MCP server.
Digital Control Manager Backend
Benchmarking zero-shot and fine-tuned time series foundation models for process model forecasting on directly-follows time series from event logs.
Implementation of Salesforce's Moirai moel
Arabian Sea SST forecasting: LSTM, N-BEATS, Moirai (55M fine-tuned) + novel 4-Stage Post-Processing Pipeline. RMSE 0.108C. INCOIS.
Unified API for 5 time series foundation models (Chronos, TimesFM, MOIRAI, Lag-Llama, TTM) with auto-select, uncertainty-weighted ensembling, and reproducible Monash benchmarks.
Backtesting popular Foundation models developed by Salesforce, Google, IBM, & Amazon that supports zero shot forecasting.
13 time-series foundation models vs classical, ML and naive baselines: 10 datasets, 4 horizons, one venv per family. Foundation models won 30 of 38 contests; the GIFT-Eval leader's open predecessor (Toto-Open-Base-1.0) won none.
To associate your repository with the moirai topic, visit your repo's landing page and select "manage topics."