Saurabh Raj Varma

I’m an early-career data and machine learning practitioner who turns messy data into reliable predictions—forecasting trends, scoring risk, and extracting signal from text. In internships, I developed time-series and classification pipelines that improved accuracy and reduced error using methods like Auto-ARIMA, LSTM, feature engineering, and cross-validation.

My work spans end-to-end modeling: EDA and benchmarking, imbalance handling, and rigorous evaluation (MAE, ROC-AUC, F1). I’ve built projects such as BI-RADS classification from mammography reports using TF-IDF, ensembles, and transformer-based models (BERT), plus churn, time-series forecasting, and fraud anomaly/risk scoring systems.

I also bring practical, results-driven thinking from real-world problem solving—expanding onboarding across Southeast Asia and designing an agro-input business model through P&L and logistics planning. I’m recognized on Kaggle as an Expert (Top 2%) and achieved a high rank in the Predict Customer Churn competition, and I enjoy pairing strong modeling with clear analysis for actionable decisions.