Skip to content
Back to work

Data platform · Forecasting

Denryoku

Electrical grid monitoring on French open data, with forecasts and a natural-language assistant.

forecast horizon
24 to 72 hforecast horizon
assistant for operational queries
RAGassistant for operational queries

The problem

RTE publishes national consumption and production data in near real time, but turning it into alerts, dashboards and forecasts an operator can question takes a full pipeline.

What I built

  1. 1

    Built real-time ingestion of eCO2mix data into time-series storage, with consumption and production indicators and threshold alerts.

  2. 2

    Designed operational dashboards in Plotly Dash.

  3. 3

    Added a LangGraph RAG assistant answering questions in natural language, with 24 to 72 hour forecasts from Darts.

  4. 4

    Shipped with GitLab CI/CD pipelines and Sphinx technical documentation.

Outcome

  • An end-to-end monitoring stack, from raw open data to forecasts an operator can query in plain language.

Have a similar problem?

Tell me about it. I reply within 48 hours with a first read of the problem and an honest opinion on fit.

Book a call