At Celebal Technologies, I developed an Estimated Time to Sell workflow for vehicle inventory using CatBoost and survival analysis. I engineered features and trained models using vehicle, market, and inventory signals.
At the Defence Research and Development Organisation (DRDO), I built an end-to-end machine learning pipeline to classify PDF files as malicious or benign, achieving 98% accuracy. I also conducted exploratory data analysis, feature engineering, and cross-validation under government security constraints.
In my projects, I created a locally hosted AI analytics assistant that connects natural-language questions to verified metrics. I also developed a solar inverter fault-detection pipeline with predictive modeling, anomaly scoring, and SHAP-based explainability.
