Projects
A showcase of my technical projects and software applications
Net Worth
Budget
An AI-powered personal finance dashboard that provides personalized financial advice and budget optimization. This project integrates OpenAI API for intelligent financial guidance and features dynamic data visualization.
Challenges:
Integrating external AI services (OpenAI API) for personalized advice generation. Developing scalable backend microservices for budget optimization. Designing an intuitive user interface for complex financial data visualization.
Solutions:
Backend API & Database: Engineered Flask RESTful APIs and a SQLite database for financial data management. AI Integration: Developed a custom prompt system using the OpenAI API to deliver personalized financial advice. Interactive Frontend: Designed and implemented a React/Next.js frontend featuring dynamic Chart.js visualizations.
Impact:
Awarded 1st Place FinTech Winner at the UAInnovate Hackathon. Provided users with an AI-powered personal finance dashboard that simplifies complex financial terms. Created a scalable platform for financial data handling and advice generation.
Technologies:
A feedforward neural network built from scratch using TensorFlow's low-level API to classify handwritten digits from the MNIST dataset. This project demonstrates deep understanding of neural network fundamentals and core machine learning algorithm implementation.
Challenges:
Implementing a neural network from fundamental components (layers, activation functions) without high-level abstractions. Manually coding the training process, including mini-batching and gradient application. Achieving competitive accuracy on a standard dataset with a custom-built model.
Solutions:
Custom Neural Network: Constructed a feedforward neural network by developing DenseLayer and SequentialLayer classes, defining weight and bias handling, and integrating activation functions. Manual Training Loop: Implemented a custom update_weights function and BatchGenerator to manage mini-batch gradient descent and the training epoch cycle. TensorFlow Core Integration: Used TensorFlow's low-level API for tensor operations, variable management, and automatic differentiation (tf.GradientTape).
Impact:
Achieved 95% accuracy on the MNIST dataset using a neural network built from scratch. Demonstrated proficiency in core machine learning algorithm implementation and low-level TensorFlow usage.
Technologies:
An interactive 3D simulation of a Rubik's cube built in Java using Processing. This project demonstrates proficiency in object-oriented programming, 3D graphics rendering, and real-time interactive systems.
Challenges:
Accurately modeling a 3D Rubik's Cube and its independent, rotatable faces. Implementing precise 3D rotational transformations for cube faces. Achieving real-time rendering performance for interactive manipulation.
Solutions:
Object-Oriented Design: Developed Cube and Block classes using OOP principles to define the cube's structure and behavior, ensuring modularity. 3D Graphics Implementation: Utilized Processing in Java with P3D mode to render the 3D cube and handle graphical transformations. Direct Control System: Implemented keyboard-based controls for immediate cube face rotations and scrambling, designed for high responsiveness.
Impact:
Built an interactive 3D Rubik's Cube simulator achieving up to 144 FPS performance. Demonstrated proficiency in OOP for complex system design and real-time graphics programming.
Technologies:
NFA
DFA
Implementation of the subset construction algorithm for converting non-deterministic finite automata (NFA) to deterministic finite automata (DFA). This project demonstrates deep understanding of theoretical computer science concepts.
Challenges:
Implementing the complex subset construction algorithm correctly, handling edge cases in automata conversion, and creating an efficient representation of the resulting DFA.
Solutions:
I carefully implemented the subset construction algorithm with proper state management, added comprehensive testing for various automata types, and optimized the DFA representation for efficient string matching.
Impact:
The implementation serves as both a learning tool for automata theory and a practical utility for regular expression processing and pattern matching applications.
Technologies:
A game engine for Mancala with an intelligent AI opponent and strategic gameplay. This project demonstrates algorithm design, game theory, and artificial intelligence implementation.
Challenges:
Creating an AI that could play Mancala strategically, implementing efficient game state evaluation, and designing an algorithm that could look ahead multiple moves.
Solutions:
I implemented the minimax algorithm with alpha-beta pruning for efficient search, developed a sophisticated evaluation function for game states, and optimized the search depth for real-time gameplay.
Impact:
The AI can consistently beat human players and demonstrates advanced strategic thinking, making it an excellent tool for studying game theory and AI algorithms.
Technologies:
A classic TicTacToe game with an intelligent AI opponent that uses the minimax algorithm to play optimally. This project demonstrates game development, AI implementation, and user interface design.
Challenges:
Implementing an unbeatable AI using the minimax algorithm, creating an intuitive user interface, and ensuring the game provides an engaging experience despite the AI's perfect play.
Solutions:
I implemented the minimax algorithm with alpha-beta pruning for optimal performance, designed a clean and responsive user interface, and added difficulty levels to make the game more accessible.
Impact:
The game serves as an excellent demonstration of AI algorithms and has been used as a teaching tool for computer science students learning about game trees and search algorithms.