This project focuses on building a robust and scalable data pipeline to manage and process internal faculty documents for use as a knowledge base for a Large Language Model (LLM). The documents come in various formats—PDF, DOCX, and JPG—each requiring different handling strategies for data extraction and it also includes a web scraping module to automatically collect relevant information from the faculty’s website, enriching the knowledge base with up-to-date content.. The main goal is to extract clean, well-structured Markdown (.md) content from these documents, preserving important information while ensuring consistency and usability for downstream LLM applications.

It uses MinIO as both a Data Lake and Data Warehouse to manage raw and processed files. To save storage, a binary patch diff strategy is applied, storing only differences from a baseline file. Inspired by DVC, it supports document versioning and traceability. The system integrates with the admin role of PYUI GPT, allowing centralized document management and updates

This project is a real-time monitoring system designed to improve the efficiency and reliability of industrial machine operations through predictive maintenance. The system collects live sensor data—such as power, pressure, voltage, force, and punch position—from manufacturing equipment and visualizes it through an interactive dashboard. It also integrates historical data and audio logs to support condition-based monitoring and troubleshooting. I implemented a real-time data streaming mechanism to receive and push live sensor readings to the dashboard, a REST API service to allow querying and interaction with stored machine logs, audio files, and system states, a control interface for connecting/disconnecting data streams, clearing logs, and triggering specific machine operations remotely.

The dashboard enables operators and engineers to: Monitor machine health in real-time Review past machine performance Detect anomalies or warning signs early Access audio logs for acoustic anomaly detection Respond to system alerts effectively The goal of the project was to support a predictive maintenance strategy—shifting from reactive to proactive machine maintenance, thereby reducing downtime and improving operational efficiency.

I was assigned to develop an automated sorting machine by using ESP32 as a microcontroller. I developed an users application to manage the use of this machine, It includes a user registration system using student IDs and a point accumulation system for users who utilize our machine. And also developed a back office application that handles about current and history score monitoring for admins. All of this, I implemented with Next.js and Firebase.