Groendus is the one-stop B2B shop for green energy, helping clients manage volatility in a decentralized grid. To optimize costs and maximize renewable usage, they need precise forecasts for everything from solar production to spot prices. Faced with an explosion of potential algorithms (consumption, generation, battery optimization), Groendus needed to decide what to buy, what to build, and how to scale without drowning in technical debt.
Outperformed external benchmark
the new customised forecasting model surpassed the accuracy of their previously purchased external forecasting provider.

The Challenge

Scaling AI in the energy sector presents unique volatility and volume issues:

  • Model Explosion: Controlling assets requires dozens of simultaneous forecasts (15-minute intervals) for consumption, solar, and price.
  • Strategic Dilemma: The “Buy vs. Build” question—when does building in-house yield a competitive advantage over buying standard vendor data?
  • Engineering Maturity: The existing data setup (standard BI) lacked the rigor (CI/CD, versioning) needed to deploy production-grade ML reliability.

Our Approach

We moved Groendus from ad-hoc analysis to a streamlined AI software company setup with the thre following steps;

  • Strategic “Buy vs. Build”: We didn’t just start coding. We analyzed where Groendus held a competitive edge. We identified that for energy consumption, their unique, granular client data allowed us to build internal models that significantly outperformed generic external market forecasts.
  • MLOps Foundation: We built the factory before the car. We implemented a robust MLOps platform (standardized training, deployment, and monitoring pipelines). This allows the team to manage a fleet of algorithms simultaneously, reducing reliance on single data scientists and ensuring 24/7 reliability for critical energy steering.
  • Analytics Engineering: To scale data delivery, we professionally engineered their data warehouse using the Medallion Architecture (Bronze/Silver/Gold). We upskilled their business analysts into Analytics Engineers, enabling them to build production-grade pipelines that serve both BI dashboards and advanced ML models from a single source of truth.
Share:

More Cases

SnappCar
New ML Ops setup delivers 50% faster queries and significant cost reduction for peer-to-peer car-sharing platform.
Crisp
Implement a new forecasting system for Crisp that improves prediction accuracy (WAPE) by >20%.
CircularPSP
Turning fragmented municipal data into actionable circular economy intelligence for policymakers across Europe.