Vairavan // Portfolio
Photo of environmental sensor hardware used for an ML experiment

Experimental Systems

Environmental Sensor and ML Experiment

An end-to-end experiment that collected temperature and humidity data from sensors, stored it in SQL, and tested whether machine learning could reduce air-conditioning usage.

RSQLPower BIArduinoStatistical Testing IoTMachine LearningEnvironmental DataEnd-to-End Workflow

Outcome: Built a full workflow from sensor hardware to analysis, showing how environmental data could support more efficient air-conditioning decisions.

Problem

Reducing air-conditioning usage sounds like a simple optimization problem until you try to measure it properly.

To do that well, I needed a workflow that could collect live environmental data, store it reliably, and make it available for both analysis and dashboarding.

Approach

This project was designed as an end-to-end system rather than a single analysis notebook.

  • Arduino-based sensors collected temperature and humidity readings.
  • The data was posted into SQL storage at frequent intervals.
  • R and Power BI were used for analysis, comparison, and reporting.
  • Statistical and predictive methods were tested to see whether air-conditioning behavior could be modelled well enough to support smarter control decisions.

What Made It Useful

The project connected several layers that are often treated separately:

  • hardware collection
  • data ingestion
  • storage design
  • analysis
  • reporting

That made it more than a modelling exercise. It was an early full-stack data system, built around a physical process instead of a static dataset.

Why It Mattered

The experiment helped show how technical systems work becomes more valuable when it stays connected to a practical decision.

In this case, the decision was energy usage. The broader lesson was that even small sensor systems benefit from being designed as complete workflows rather than isolated prototypes.

Back to all technical work