AI Engineer | 2026 ASTRI Intern | Final Semester MSc AI (CUHK) | CompTIA Security+ | RPI Alum

Python | PyTorch | OpenCV | Computer Vision | Deep Learning | Multimodal AI

Building multimodal AI pipelines

MSc AI candidate at CUHK focused on computer vision, multimodal AI, and pipeline development. Built a 6-stage multimodal pipeline at ASTRI from scratch — integrating YOLOv8, ResNet-50, custom OCR, and Qwen-VL — learning on the job and delivering a working prototype with 70–120% performance improvement.

Background includes security (RPI, CompTIA Security+, NCL Top 5%) and full-stack development. Now looking to apply my AI skills to build programs that automate tasks and get more precise results.

Seeking: Full-time AI Engineer / ML Engineer roles in computer vision, multimodal AI, or AI/ML engineering.

Core AI Computer Vision Deep Learning Multimodal AI

ML Engineering PyTorch OpenCV YOLO OCR CVAT Model Training Data Annotation

Languages Python JavaScript C

Development React Node.js Git

Security Vulnerability Assessment Cryptography

Cloud & Infrastructure AWS Docker MongoDB

AI Memory Systems & Game Integration - Local AI Dungeon Master with persistent memory across gaming sessions

• Role: Prompt engineering and vector database configuration
• Impact: Designed world-building templates and implemented ChromaDB memory storage for consistent AI personalities
• Tech: mem0ai, Ollama (Qwen3:8B), ChromaDB, Neo4j, Python 3.12+
• Note: Research project exploring persistent memory systems for D&D gameplay

Model Architecture & Benchmark Analysis - Comparative survey of six VLMs analyzing both design and trade-offs

• Role: Architecture analysis, benchmark evaluation (MME), and model-specific study (InternVL, Mini-Gemini)
• Impact: Authored A-grade comparative survey paper for graduate course
• Tech: Literature review, model architecture analysis, benchmark comparison
• Note: Co-authored for graduate course at CUHK; full paper available upon request

Systems Security - Recreated Ken Thompson's seminal attack by modifying the TCC compiler

• Role: Architecture analysis and exploit development based on published concept
• Impact: Authored comprehensive methodology and validated self-replicating backdoor functionality
• Tech: C, compiler internals (libtcc)
• Note: Implementation completed based on Thompson's theoretical description before his specific technical approach was published

Security Architecture - Analyzed architectural vulnerabilities in cloud-connected IoT systems

• Role: Threat modeling and security analysis
• Impact: Authored A-grade research paper proposing defense models
• Tech: Threat modeling, cloud security
• Note: Research conducted collaboratively; security analysis and paper authored individually

Cryptography & Network Security - Two-phase educational project implementing defensive protocol design and offensive security assessment

• Role: Paired programming for protocol implementation and security assessment
• Impact: Uncovered critical flaws including weak key space and hardcoded credentials in peer implementations
• Tech: Python, RSA, S-DES, HMAC-SHA1, cryptanalysis

Industry Application - Full-stack system using computer vision to automate aircraft safety compliance checks

• Role: Frontend development and user interface design
• Impact: Built and iteratively improved user interface based on stakeholder feedback, enabling successful demonstration
• Tech: JavaScript, HTML, CSS, Bootstrap, UI/UX design

Web Platform - Public website and CTF management system for RPI's cybersecurity lab with separate frontend/backend services

• Role: Backend API and database architecture
• Impact: Delivered functional backend service enabling student training and write-up management
• Tech: Node.js, Express.js, MongoDB, RESTful API design

I'm always open to discussing:

• Multimodal AI and computer vision projects
• AI/ML engineering and research collaboration
• Security challenges in AI and software systems

Reach out via LinkedIn or Email: raphael.chung.sh@gmail.com