ClevAir delivers a smart system that manages the building’s energy consumption, automates its maintenance and operations, revitalizes its climate, and offers you all the insights you need to run it even better. The system enables reducing energy costs and minimizing a negative impact on the environment.
Meet Our Client
ClevAir delivers a smart system that manages the building’s energy consumption, automates its maintenance and operations, revitalizes its climate, and offers you all the insights you need to run it even better. The system enables reducing energy costs and minimizing a negative impact on the environment.
Case Study Shortcut
Challenge
Manual Data Labeling Slowed Down Onboarding and Scalability
ClevAir’s data labeling process was entirely manual, requiring engineers to clean and annotate large volumes of data from scratch every time a new client was onboarded. This significantly slowed down customer acquisition and limited the company’s ability to scale efficiently.
Disordered and Unlabeled Data from Heterogeneous Sources
The company collected a wide range of sensor and system data in various formats and quality levels, often inconsistent, unlabeled, or containing errors. The lack of standardized, labeled data made automation and real-time optimization challenging.
Lack of Automated Data Type Detection
ClevAir had no system in place to automatically detect and classify semantic data types. This meant every dataset required manual interpretation and transformation rules, resulting in a time-consuming and error-prone preprocessing pipeline.
Goal
- Automate the Data Cleaning and Labeling Process
- Enable Fast and Scalable Client Onboarding
- Improve Data Quality with Semantic Type Detection
Outcome
Now the company’s software is able to automatically detect data types and transform data accordingly, for example using a map to visualize the value pairs.
With automation, the company is able to speed up client acquisition processes as pre-optimization tasks of collecting data go smooth and error-free. New client doesn’t have to wait for the business insights and knows from the very beginning the value they can get.
Before
- Manual data cleaning and labeling process
- Slow business growth and lack of scalability
After
- Automatic data cleaning process
- Faster acquisition of new customers
- Business scalability
Integrate those solutions in your company
Case Study Details
Approach
Automated Semantic Data Type Detection
- Developed a system to automatically detect the data type of each data point in diverse datasets.
- Replaced manual rules with a machine learning model trained on real-world semantic examples, capable of understanding context (e.g., dates, locations, numerical values).
Model Robust to Dirty and Unstructured Data
- The model was designed to be resilient to noise, inconsistencies, and formatting issues, common in data collected from smart buildings and IoT sensors.
- This made it well-suited for real-world energy datasets, which are often incomplete or messy.
Predictive Performance Over Traditional Rule-Based Systems
- Traditional decision tree models were evaluated but discarded due to limitations in predictive power and scalability.
- The chosen model outperformed custom rule sets, offering better accuracy and greater adaptability to new datasets.
Built for Scalability and Deployment in AWS Environment
- The solution was implemented and tested in an AWS-based environment, ensuring cloud readiness and easy deployment across client systems.
- Designed for scalable integration, enabling ClevAir to apply the system seamlessly during new client onboarding.
