Machine Learning
DataCool
Python • ML • Streamlit
Role
ML Engineer
Timeline
2025
Tools
Python, scikit-learn, Streamlit, SHAP, Plotly
The Problem
Data centers face thermal management challenges. Server hotspots reduce efficiency, increase cooling costs, and risk hardware failure. Traditional monitoring is reactive, not predictive.
The Solution
Built a machine learning system using Histogram Gradient Boosting to predict hotspots with 94% accuracy. Implemented constraint-based optimization for workload redistribution with thermal physics modeling.
Results
- 0194% prediction accuracy on hotspot detection
- 0275% reduction in critical overheating incidents
- 03Interactive 3D visualization dashboard