Vendor Invoice Intelligence Portal
96.99% R² · 94% acc
An end-to-end dual-model financial risk application combining freight cost regression (R² = 96.99%) and an invoice risk classification engine (94% accuracy). Features automated hyperparameter optimization via GridSearchCV, an embedded SQLite relational schema for sub-second queries, and a live Streamlit interface for finance teams.
Scikit-learnSQLiteGridSearchCVStreamlit CloudPandasDual-Model Architecture
Career Mentor Agent
Autonomous Agent System
An autonomous conversational mentor developed during the Microsoft Frontier Agent-a-Thon. Built using Python, the GitHub Copilot SDK, and AsyncIO runtime loops. Features dynamic tool routing, customized career roadmapping, technical interview scenario generation, and persistent multi-turn session memory deployed on Streamlit Cloud.
GitHub Copilot SDKPythonAsyncIOAutonomous AgentsStreamlit CloudPrompt Engineering
Financial Lead Scoring & SHAP
91% ROC-AUC
Benchmarked Random Forest against Logistic Regression classifiers for bank term deposits, applying SHAP values to explain feature contributions (call duration, account balance, age) to financial decision-makers.
Scikit-learnSHAP (Explainable AI)Random ForestLogistic Regression
Customer Segmentation Engine
K-Means · PCA
Engineered an unsupervised customer clustering pipeline using PCA dimensionality reduction and K-Means to identify and visualize 5 actionable demographic and spending profiles.
K-MeansPCADimensionality ReductionMatplotlib
E-Commerce BI Intelligence
Power BI Dashboard
Interactive business intelligence report analyzing transaction patterns, profit margins, and cohort performance designed for non-technical executive decision-making.
Power BIDAXData ModelingSQL
Bookstore Relational SQL Engine
MySQL · 8 CTEs
Normalized relational schema with 500+ records demonstrating window functions, CTEs, running revenue totals, and complex joins for revenue trends and inventory audits.
MySQLWindow FunctionsCTEsRelational Design
