CAPSTONE 3D UCSB — Campus-Scale 3D Reconstruction
Ongoing
Built a campus-scale 3D reconstruction pipeline using SfM and Gaussian Splatting to produce an interactive digital twin of UCSB. Implemented privacy-aware filtering using YOLO, Laplacian sharpness analysis, and metadata consistency checks. Developed incremental COLMAP/HLOC workflows for multi-session reconstruction, point cloud merging, and building expansion. Integrated DISK + RoMa feature matching to improve robustness under wide-baseline views and outdoor lighting. Optimized dense reconstruction with Gaussian Splatting using adaptive density control for efficient GPU rendering.
Python
COLMAP
Gaussian Splatting
YOLO
Computer Vision
PC Builder Agent
Recent
Built an AI agent that translates workload and budget constraints into validated PC build recommendations. Implemented a plan-validate-revise loop with generative planning and deterministic compatibility checks to ensure feasible configs. Designed modular architecture separating reasoning, validation, and explanation to improve reliability and extensibility.
Python
LLMs (OpenAI / Ollama)
Pydantic
AI Agents
Air Traffic Simulation
2024
Multi-agent simulation system for autonomous delivery drones in urban environments. Implements reinforcement learning for optimal flight paths, LSTM models for demand forecasting, and dynamic fleet management with battery tracking and charging stations. Features real-time visualization with Pygame and environmental factors affecting drone performance. Achieved 22% improvement in operational efficiency through intelligent route planning.
Python
TensorFlow
Q-Learning
LSTM
Pygame
Reinforcement Learning
Housing Market Analysis & Forecasting
2024
Analyzes historical housing and job market data to forecast future affordability and suggest optimal living locations. Employs ARIMA models for price forecasting and Random Forest regressors for appreciation prediction. Features affordability scoring and multi-criteria location ranking with interactive visualizations using matplotlib and seaborn.
Python
Pandas
scikit-learn
ARIMA
Random Forest
Data Analysis
Pac-Man on Arduino
ECE 5 - Winter 2024
Implemented classic Pac-Man game on Arduino Mega with TFT touchscreen display. Features intelligent ghost AI using BFS pathfinding, power-up mechanics, score tracking, and joystick controls. Built with modular architecture including separate classes for actors, ghosts, game state, and controller management.
C/C++
Arduino
Embedded Systems
BFS Algorithm
Game Development
Data Structures & Algorithms (CS24)
Spring 2023
Comprehensive implementation of fundamental data structures and algorithms in C++. Includes nine programming assignments covering FibVec, TicTac, Setree, GenePool, Counter, Typo, WordHop, and Lineup. Topics range from memory management and data structures to advanced concepts like hash tables, heaps, and graph algorithms.
C++
Data Structures
Algorithms
Python
Bloom Filter Implementation
CS130A Project
Probabilistic data structure implementation demonstrating efficient set membership testing. Uses multiple hash functions and bit arrays to store and query elements. Guarantees no false negatives while accepting occasional false positives, making it ideal for space-efficient applications where approximate answers are acceptable.
Python
Algorithms
Data Structures
Hash Functions
Stock Website Scraper
2024
Python web scraper designed to efficiently track and reflect changes in locally saved stock prices. Maintains historical records with time-based comparison (daily and monthly data) enabling price trend analysis and change detection for financial data monitoring.
Python
Web Scraping
JSON
Data Processing
Python
C/C++
JavaScript
SQL
React
Node.js
HuggingFace
TensorFlow
AWS
Azure
Google Cloud
Docker
Kubernetes
GitHub Actions
Git/GitHub
Postman