An enterprise solution to digital pollution via project optimization.
Klean was actually inspired by the understanding of Fast Fashion; and industry wide environmental problem due to consumer carelessness. Just like how people will rapidly purchase and discard items without care, we do the exact same thing online but with our data. Klean was born to bring awareness to Digital Pollution, where excess data leads to a sustainable amount of unnecessary emission and energy expenditure.
Did you know?
- 100 emails pollutes as much as 125 light bulbs
- Data centres currently account for between 1.8% and 2.8% of the world's electricity consumption
- Our digital activity pollutes more than some countries like Switzerland or Norway
![[Figure 1]](https://i.imgur.com/5Xxf2ed.jpeg)
Klean is a digital waste cleaning web app that offers Code Optimization along with Data Analysis for emissions. By analyzing and optimizing Python code to improve its efficiency and sustainability, we aim to reduce excess computational movement at an enterprise level to save energy. By leveraging machine learning models and sustainability tracking tools such as CodeCarbon, Klean identifies memory-intensive operations, redundant computations, and suboptimal data structures, and suggests improvements. The project's key achievements include significant reductions in energy consumption and carbon footprint.
![[Figure 2]](https://i.imgur.com/XXJ0a0G.png)
AI-Powered Refactoring
Base Model: Utilize Microsoft's CodeBERT as the foundational model.
Fine-Tuning: Fine-tune the model on a custom dataset of energy-efficient Python code.
Optimization Focus: Emphasize loop efficiency, memory usage, and I/O operations.
Personalization: Incorporate data from the repository owner's GitHub profile to tailor optimizations.
Iterative Optimization: Conduct multiple passes to refine and improve code efficiency continuously. The model identifies optimization opportunities by analyzing the code structure, usage patterns, and comparing them with best practices in energy-efficient coding.
- Data Set Quality: One of the main challenges was obtaining a high-quality dataset for fine-tuning the AI model. Energy-efficient code samples are not widely available, which required extensive research and collaboration with experienced developers to curate a suitable dataset.
- Complexity of Code Structures: Optimizing code that is not only energy-efficient but also maintains its functionality and performance posed a significant challenge. It required a delicate balance between refactoring code and preserving the original intent and output of complex algorithms.
- Model Accuracy and Performance: Ensuring that the AI-powered refactoring tool accurately identifies and suggests optimizations without introducing errors or inefficiencies was a key challenge. This required iterative testing and refinement of the model.
- Integration with Existing Systems: Incorporating Klean into existing enterprise systems required addressing compatibility issues and ensuring that the tool could seamlessly integrate with various coding environments and workflows.
Our data shows the effectiveness of Klean in optimizing computational tasks for better efficiency and sustainability. Klean's approach consistently improves energy consumption, reduces carbon footprints, and enhances execution times across various metrics. These optimizations contribute significantly to environmental sustainability, and beat out the competition by minimizing digital waste and its associated impacts, demonstrating Klean's potential as a powerful tool in the field of sustainable computing.
Klean currently primarily focuses on large scale code optimization, but in the future should branch into:
Data Transferring and Cache Clearing: By allowing enterprises to clear unused temp files and bandwidth at a large scale, companies are able to save significantly on database costs and reduce their carbon footprint
User-centered Cloud Optimization: By setting up user-sided data management software, it empowers users to take initiative of their own digital waste and clear up their useless files, photos, messages, etc.
Automatic Data Auditing: By automating digital clean-up, Klean may have the opportunity to become a scalable product, ready to be released into the market and monetized.
Government incentivized Mandatory Audits: Through collaboration with government sustainability concerns, Klean has the potential to become mandated/enforced by regulations to reduce digital waste at a much larger scale. This will aid in the promotion of Klean as a software product, along with the general incentive for large corporations to be cognizant of their digital waste.
