Marcus Valente Software development
All projects

Kairos

Machine learning to classify fleet maintenance risks using operational data.

Context
Academic project · Inteli + Fadel Transportes
My role
Data cleaning and feature engineering
Period
August – October 2025
In collaboration with
Fadel Transportes

The challenge

Kairos used fleet history, odometer readings and operational attributes to predict corrective maintenance needs over the next 30 days. The team's solution used Random Forest and connected the analysis to a Streamlit dashboard.

My contribution

I was responsible for data preprocessing, including cleaning, and feature engineering. The work aimed to improve predictive-model metrics by preparing the data and variables used in the analysis.

The delivery

The team developed the academic project with Fadel Transportes between August and October 2025. My delivery focused on the prepared dataset and features used by the model.

Scope and limits

Improved metrics were the objective of data preparation. This case does not attribute numerical gains to my work or claim measured reductions in failures or maintenance costs.

From history to model

Preparing the dataset

My work started with processing the data before it was used by the model.

Building features

I worked on feature engineering using the project data to improve the inputs available for prediction.

Team prediction and visualization

Random Forest and Streamlit formed the solution for risk classification and consulting fleet analysis.

Technologies used

  • Python
  • Pandas
  • NumPy
  • Scikit-learn
  • Streamlit
Next projectInsPecT

contact.md

Shall we talk?

For software engineering opportunities and internships, with room to contribute through applied AI and data. I reply within 3 days.