Preethin Peter

I’m a data analyst and GenAI-focused data science graduate who builds predictive models and end-to-end pipelines that drive real business decisions. My work blends machine learning, forecasting, and LLM tooling to turn complex data into clear, action-ready outcomes.

In my Data Science Intern role at Wise Analytics, I built an end-to-end AI pipeline for food spoilage prediction and inventory optimization across retail stores. Using Random Forest on 100K+ grocery records across 10 stores, I achieved ROC-AUC of 0.96 and R² of 0.98 for early detection of high-risk perishable stock.

I also developed a demand forecasting pipeline using Prophet and XGBoost, producing 30-day store-level forecasts (avg. MAE: 134 units) through time series decomposition and seasonality analysis to reduce stockouts. To make recommendations more usable, I engineered a LangChain multi-agent system with 4 AI agents (Inventory, Demand, WeatherImpact, Decision) processing 29,500+ transactions and generating explainable natural language recommendations per store.

To balance growth with cost control, I designed 3 PuLP linear programming models for discount allocation, replenishment planning, and surplus redistribution across 10 stores—optimizing waste minimization while protecting revenue. I bring a strong bias for statistical rigor, data quality, and communication that helps stakeholders move faster.