CloudML: Algorithms to track your server cost

For my B.S. thesis, I developed CloudML, a platform designed to help businesses predict cloud service costs and potential failures using machine learning. This project was a collaboration with XalDigital, they provided me the training data and I developed the data pipeline, the trained models and the platform.

This project was the culmination of everything I had learned during my Computer Science degree, combining AI, full-stack development, and real-world problem-solving.

Turning Data into Actionable Insights

The goal of CloudML was to give users the tools to make informed decisions about their cloud infrastructure. Here’s what I accomplished:

Predicting Failures and Costs with AI

Using advanced machine learning models, I achieved 72% accuracy in predicting system failures and an impressive error rate of just $0.02 when forecasting future service costs. These predictions empowered users to optimize their cloud resources and avoid unexpected expenses or downtime.

How to use it

First you to have your data in .csv format, it will be parsed given each of the current processes you have running in your server, then in the background a machine learning model will be trained and when it finished you will be received the performance analysis of the final model. Then you can use the model to predict future costs and failed processes using the same data format excluding costs.

Final Steps

Working on CloudML taught me the importance of integrating AI with user-friendly tools to solve real-world challenges. I gained hands-on experience in full-stack development, model deployment, and optimizing system performance. This project solidified my passion for combining machine learning with practical applications, showing me how AI can make complex tasks more accessible and efficient.