Architected a secure MERN backend with JWT-based authentication and role-based authorization. Engineered a robust CI/CD pipeline using GitHub Actions to automate ESLint v9 quality gates and asynchronous integration testing, ensuring 100% build reliability and deterministic state management in a production-ready environment.

Developed an end-to-end risk assessment system combining a Random Forest Classifier with Retrieval-Augmented Generation (RAG). Solved the 'black box' problem by using ChromaDB and LangChain to provide transparent, policy-backed justifications for loan decisions.

Developed a high-performance Model Context Protocol (MCP) server to bridge LLMs with real-time GitHub data. Engineered custom repository health metrics using Zod-based strict validation and parallel API orchestration to provide AI models with feature-engineered insights.

Developed a centralized middleware server to aggregate real-time shipping rates from multiple logistics providers (J&T Express, City-Link). Implemented the Strategy Pattern to decouple provider-specific logic, integrated SpringDoc OpenAPI for automated Swagger documentation, and containerized the service using a multi-stage Docker build for microservice compatibility.

Developed a full-stack task management system with an Angular frontend and a Node.js/Express backend. Implemented RESTful APIs for CRUD operations and integrated MongoDB Atlas for data persistence, handling cross-origin resource sharing and production deployment.

Achieved a top-tier RMSE of 0.12230 by implementing a multi-stage Stacked CV Regressor. Optimized the bias-variance tradeoff using log-transformations, skewness correction, and an ensemble of Lasso, Ridge, and LightGBM models.

Engineered a predictive pipeline to identify high-risk telecom customers. Mitigated a 27% class imbalance using SMOTE and optimized a Logistic Regression model via threshold tuning to achieve a 79% Recall score.

Developed a Random Forest Classifier to predict survival outcomes. Engineered features from passenger names to identify titles and family units, capturing historical patterns of the 1912 disaster.

Completed RAG-based chatbot projects using LangChain and LLMs in Python. Tuned temperature and prompt structures to optimize context-aware responses.

Architected an autonomous agentic workflow using n8n to deliver personalized daily briefings. Integrated Google Gemini (LLM) with real-time OpenWeatherMap API data to generate context-aware insights, utilizing OAuth2 for secure Gmail delivery and automated cron-based orchestration.

Built a desktop application with SQL integration and booking validation for a localized movie theater system.

Developed an arcade-style game using PyGame featuring collision detection, event-driven player controls, and finite game state management.