Skip to content
Pratim/Dasude
← All experience

02Jul 2023 - Nov 2023

Data Science Intern

Highradius Technologies

Developed deduction classification models with XGBoost and LightGBM, improved accuracy through feature engineering, and automated monitoring pipelines to reduce drift-driven prediction failures.

01What I did

  1. 01

    Classification modelling

    Developed gradient-boosted models (XGBoost, LightGBM) to classify payment deductions, comparing them against baselines to pick the best performer.

  2. 02

    Feature engineering

    Designed features from transaction and customer history that lifted model accuracy over the raw-column baseline.

  3. 03

    Monitoring automation

    Automated monitoring pipelines that watch input and prediction distributions, reducing failures caused by data drift.

02Impact

  • Higher classification accuracy from engineered features
  • Fewer drift-driven prediction failures through automated monitoring

03Tech

XGBoost, LightGBM, Python, Feature engineering, Drift monitoring