Introduction

Gary Phua

Gary Phua

Data Scientist & AI Engineer

šŸ“ Singapore

My name is Gary and the website you see along with the relevant projects are built with the sole purpose for me to join Rocket Lab. Humans only learnt to fly barely a lifetime ago. Since then, we have looked to the stars, and gone from flying in the atmosphere to landing on the Moon in less than a lifetime. I have long been interested in space and science fiction, and in a way what Rocket Lab is doing is pushing science fiction to reality, in terms of both developing space systems and orbital launch capabilities. I hope to be a part of that future if given a chance. I also believe in doing and not just words. This page contains introductions and links to two projects that have been deployed. The first project is a rocket trajectory simulator that i built with electrons specifications, and the second was built after a discussion with Otso from Rocket Lab about applying AI to space systems. I trained a vision transformer model on the DAGM dataset to achieve ~80% accuracy on surface defect detection. With my background in both physics and AI, as well as implementing these projects, I hope to have demonstrated my capability and be given a chance to help contribute to Rocket Lab and its mission. To go to space to make life better on Earth.

Education

  • PhD Computer Science Singapore University of Technology and Design President's Graduate Fellowship (2016-2021)
  • BSc Physics (First Class Honours) Nanyang Technological University (2011-2015)

Core Expertise

Python Machine Learning Computer Vision PyTorch NLP Data Science SQL LangGraph Generative AI

Space-Grade Engineering Projects

01

Electron Rocket Trajectory Simulator

šŸš€ LIVE DEMO

High-fidelity orbital trajectory simulation for Rocket Lab's Electron vehicle. Demonstrates actual flight dynamics from launch to orbit insertion using real specifications.

Global Launch Sites
0-320kg Payload Range
3D Visualization
Real-time Telemetry

Technical Implementation

Physics Model
  • N-body gravitational dynamics with J2 perturbation
  • Exponential atmospheric density model (ρ = ρ₀e^(-h/H))
  • Aerodynamic drag: F_d = -½ρv²C_dA
  • Earth rotation effects (Coriolis and centrifugal)
Electron Specifications
  • Stage 1: 9Ɨ Rutherford engines, 183kN thrust, ISP 311s
  • Stage 2: 1Ɨ Rutherford Vacuum, 25.8kN thrust, ISP 343s
  • Propellant: 9,300kg (S1) + 2,150kg (S2)
  • Adaptive pitch program for gravity turn
Numerical Methods
  • 4th-order Runge-Kutta integration (RK4)
  • Adaptive timestep for critical phases
  • Coordinate transformations (ECI/ECEF/LVLH)
  • Orbital element calculation (semi-major axis, eccentricity)
Intuitive User Interface
  • Interactive 3D Earth with real-time trajectory rendering
  • Adjustable simulation speed (0.1x - 10x)
  • Dynamic payload mass configuration (0-320kg)
  • Global launch site selection with accurate coordinates
  • Live telemetry dashboard with altitude, velocity, and mass
  • Stage separation visualization and tracking
02

Material Defect Detection System

šŸš€ LIVE DEMO

Advanced Vision Transformer model for real-time defect detection in aerospace materials. Critical for ensuring structural integrity and quality control in space systems manufacturing.

97.02% Total Accuracy
79.52% Defect Detection
0.8743 F1 Score
1s Inference Time

Technical Implementation

Model Architecture
  • Vision Transformer (ViT) with efficient attention mechanisms
  • Trained on NVIDIA RTX 3090 with FP32 precision
  • Patch size: 16x16, Image size: 224x224
  • Multi-scale feature extraction for defect patterns
Training Pipeline
  • Dataset: DAGM 2007 defect detection benchmark
  • Advanced augmentation: rotation, noise injection, brightness/contrast
  • AdamW optimizer with cosine annealing scheduler
  • Automatic mixed precision training for efficiency
Deployment Features
  • FastAPI backend with async processing
  • Streamlit web interface with Rocket Lab theme
  • Real-time inference with GPU acceleration
  • RESTful API for system integration
Space Applications
  • Carbon composite inspection for rocket structures
  • Weld quality verification for propellant tanks
  • Surface defect detection in thermal protection systems
  • Automated QA for production line components

RKLB phone Gary ā˜Žļø