AI Engineer with a background in Electrical Engineering and a strong
focus on building intelligent, modular applications. Experienced in
computer vision using PyTorch, TensorFlow, and CUDA. Currently
exploring LLM-powered AI agents, with emphasis on memory systems,
modular architecture, and automation workflows. Passionate about
multi-agent systems and Web3 as emerging frontiers in intelligent
computing.
An intelligent Discord bot powered by Google Gemini, featuring
AI-driven conversations, smart message processing, automatic table
rendering, and Google Search integration. Built for personalized
user interactions with multi-language support and advanced context
awareness.
PythonDiscordLLM APIs
Post-Quantum Cryptographic Encrypted File Authorization Management
System Based on Sui Blockchain and Decentralized Storage
Developed a post-quantum cryptographic file authorization system
integrating Sui blockchain and Arweave permanent storage.
Implemented three-layer architecture with CRYSTALS-Kyber/AES-256-GCM
hybrid encryption for post-quantum security, NFT-based access
control, and immutable usage tracking. Demonstrated enhanced
security guarantees and permanent data availability through CIFAR-10
model weight validation.
Sui BlockchainArweaveCRYSTALS-KyberAES-256-GCMNFTSmart Contracts
iOS LLM Integration App
2025 (Ongoing) • iOS Development
Developing an iOS application combining LLM APIs with iOS Shortcuts
for on-device task automation.
SwiftiOSLLM APIs
LLM-powered Discord Bot with Memory
2025 • AI Agent Development
Built Discord bot integrating Groq and Gemini LLM APIs with Mem0 for
persistent conversational memory and context-aware responses.
Implemented Mem0 with Qdrant vector store for efficient memory
retrieval and Gemini embeddings for semantic understanding, enabling
the bot to maintain contextual awareness across multiple
conversations.
PythonLLM APIsMem0Discord Bot
CAE Auto-Colorization Model
2023 • Deep Learning Project
Built Convolutional Autoencoder with residual connections using
TensorFlow for automatic image colorization. Integrated ResNet-based
encoder architecture to preserve fine-grained features during
downsampling. Implemented GPU-accelerated training pipeline with
mixed precision optimization.
TensorFlowPythonCUDAImage Processing
Zero-Contact Human-Machine Interaction Solution for Epidemic
Prevention
2022 • Honorable Mention, National University Industry-Academia
Collaboration Innovation Competition
Developed gesture control system using OpenCV and MediaPipe for
real-time hand tracking and gesture recognition. Enabled computer
interface interaction through vision-based hand gestures.
PythonOpenCVMediaPipeComputer Vision
Holtek MCU Programming
2021 • Self-Driven MCU Development
Wrote full Holtek assembly language routines for microcontroller
control tasks using documentation and self-study. Shared resources
and explained usage to peers.
AssemblyMCULow-level Programming
Work Experience
Seeking opportunities to apply my AI/ML and blockchain expertise
in innovative applications, focusing on machine learning
engineering, AI agent development, computer vision, and blockchain
technologies.
During a camp I attended, there was a
challenge where we had to work with teammates to build a tower using
spaghetti and tape, with a marshmallow placed on top.My teammates
and I constructed a pentahedron within the time limit - with a
square base and four equilateral triangles forming the upper part -
and used this as the foundation to extend spaghetti vertically
upward from the center.
Everything seemed perfect, but when we
inserted the marshmallow, the tower started to bend. I realized this
wouldn’t work, so I applied tape to one side to prevent it from
falling over, though it remained bent.I initially thought other
teams might surpass us, but it turned out we were the only team that
successfully built a standing tower. This made me feel very honored.
It was only after placing the marshmallow
that I realized I had forgotten to consider its weight - this seems
to be a common blind spot in this challenge. Fortunately, our
structure was sturdy enough to keep it from collapsing. This
experience taught me not to focus too heavily on theory, but to
actually try building and make gradual corrections, and that people
can easily have blind spots.