Carbon pilot

Inspiration

What it does

How we built it

Challenges we ran into## Inspiration

Cloud computing quietly emits tons of CO₂, yet most teams have no idea how much their workloads emit. We wanted a way to make cloud carbon visible and actionable.

What it does

CarbonPilot forecasts tomorrow's energy use and carbon footprint of Azure workloads and recommends how to cut it:

• Shut down idle VMs
• Right-size over-provisioned resources
• Shift flexible jobs to cleaner-grid hours

A dashboard shows a single carbon score, forecast trends, top polluting resources and estimated CO₂ and cost savings.

How we built it

• Azure Monitor collects CPU, memory and network telemetry
• Azure Functions process the data into Azure Storage
• A [XGBoost/Prophet] model on Azure Machine Learning forecasts usage
• Energy is estimated from CPU utilization, then converted to CO₂ using regional grid carbon intensity
• [Power BI / web] dashboard shows results and recommendations

Challenges we ran into

• [Getting realistic telemetry data]
• [Estimating energy from utilization without direct power readings]
• [Time limits]

Accomplishments we're proud of

[Forecast accuracy, e.g. X% error] and [recommendations that cut emissions by X% in our tests].

What we learned

How to turn raw cloud metrics into carbon insights, and how small scheduling changes can reduce emissions.

What's next

Multi-cloud support (AWS, GCP), automatic remediation, and team-level carbon budgets.

Accomplishments that we're proud of

What we learned

What's next for Carbon pilot

Built With

• azure
• azure-blob-storage
• azure-functions
• azure-monitor
• power-bi
• python
• scikit-learn
• xgboost