How Float Runs an AI Energy Company on a 3-Person Team with Tiger Data

2026-07-08 · Source: HackerNoon · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics, Cloud Computing & IT Infrastructure · Depth: Intermediate, medium

Summary

Danish startup Float, co-founded by Jens Brandt Nellegaard and Victor Grabow in 2022, operates an AI energy company with a three-person team by achieving 99.3% compression on 1Hz smart meter data. This high compression, exceeding their 90% viability threshold, enables real-time appliance-level energy analytics for hundreds of homes, starting with a private beta for 350 pre-vetted customers. Float's system uses a proprietary hardware module, a signal processing and neural net pipeline, and a consumer app to disaggregate 15 measurements per second per household. The company, which secured its energy provider license in December 2025 after a proof of concept in December 2024, relies on Tiger Data's managed TimescaleDB for storage and continuous aggregates. This architecture supports their flat-rate subscription model, full data retention for ML training, and real-time billing, addressing a market lacking innovation and facing new consumer protection regulation by January 2026.

Key takeaway

For AI Engineers or MLOps Engineers building real-time data platforms with high-volume time-series data, you should prioritize database solutions offering extreme compression and continuous aggregation. This approach, exemplified by Float's 99.3% compression on Tiger Data, directly impacts your unit economics and operational efficiency, allowing full data retention for ML training and eliminating complex batch infrastructure. Evaluate managed services to offload database operations, enabling your small team to focus on core product development and scale effectively.

Key insights

Achieving extreme data compression is critical for scaling real-time energy analytics and enabling novel business models.

Principles

• High data compression (e.g., >90%) can make high-resolution time-series data economically viable.
• Managed database services reduce operational overhead for small teams.
• Continuous aggregates enable real-time billing and anomaly detection without batch jobs.

Method

Float's system collects 1Hz smart meter data via an IoT module, streams it through Azure IoT Hub and Google Cloud, then stores and processes it in Tiger Data using continuous aggregates for disaggregation and billing.

In practice

• Evaluate time-series databases for compression and continuous aggregation features.
• Retain raw 1Hz data for ML model training on appliance signatures.
• Use 1Hz resolution for accurate appliance differentiation.

Topics

• Time-Series Databases
• Data Compression
• Energy Disaggregation
• Smart Meters
• Tiger Data
• MLOps
• IoT Data

Best for: CTO, VP of Engineering/Data, AI Architect, AI Engineer, MLOps Engineer, Director of AI/ML

Related on AIssential

• I Spent Years in Data. It Took the Energy Industry to Humble Me.[Part -II] — Energy industry data systems are complex, real-time, and critical for operations, regulation, and revenue, demanding specialized data engineering.
• Ora Computing raises €3.5M to build the efficiency layer of the AI stack — Efficient, compact AI models, not just larger ones, will drive the next wave of AI adoption.
• Scaling for MHHS: how Octopus Energy achieved a 50x cost reduction in margin data engineering — Re-architecting data pipelines for granular processing and incremental updates drastically reduces costs and improves data freshness.
• Collecting robot training data is dirty, unglamorous work. Some AI labs are already paying XDOF to do it. — Robotics AI development is bottlenecked by scarce physical interaction data, driving demand for specialized data collection and annotation infrastructure.
• Training Time Prediction for Mixed Precision-based Distributed Training — Precision settings significantly impact distributed deep learning training times, necessitating precision-aware prediction models.
• Making User-Sequence Data More Cost-Efficient, Faster, and Easier to Use — A unified definition for user sequences ensures consistency across real-time indexing, batch processing, and online serving.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by HackerNoon.