AI & Full-Stack Engineer | Computer Vision Specialist
Bridging the gap between deep learning models and real-world applications.
I am an AI and Full-Stack Engineer with 7+ years of experience rooted in Image Signal Processing (ISP) and modern Deep Learning. I don't just build models; I own the entire project lifecycle—from architecting robust backends to deploying edge-optimized solutions.
- 🔭 Currently Working On: Developing practical AI solutions for commercial security and healthcare.
- 🌱 Learning: Advanced frontend architectures (Vue 3 composition API) and scalable infrastructure.
- 🎯 Goal: To build complete, robust, and deployable solutions from concept to completion.
- 🇯🇵 I play so many Japanese video games (like Dream Club) that I actually learned to read and speak Japanese!
- 🎰 Gacha Gamer: The rates are salty, but I keep drawing! (I even built simulators to test my luck).
- 🐍 Automation: I wrote a Python script to automatically organize my entire file system because manual sorting is for mortals.
- Core Frameworks: PyTorch, Ultralytics (YOLO), TensorFlow, Keras.
- Libraries: OpenCV, Supervision, NumPy, Scikit-learn.
- Tools: Jupyter Notebooks, Anaconda.
- Backend: Django, FastAPI, Node.js, WebSockets.
- Frontend: Vue.js, Vite, Tailwind CSS, Bootstrap, HTML5, CSS3.
- Data Visualization: Chart.js, ApexCharts.
- Databases: PostgreSQL, MySQL, MariaDB, SQLite.
- Caching & ORM: Redis, SQLAlchemy.
- Primary: Python, TypeScript, JavaScript.
- Previous Experience: C#, Java, MATLAB.
- Deployment: Docker, Nginx, Raspberry Pi (Edge Computing).
- CI/CD: GitLab CI/CD.
- Tools: Git, GitHub, VSCode, Postman, n8n (Workflow Automation).
- Core Concepts: OOP, Multi-threading, RESTful APIs, Microservices.
| Project | Description |
|---|---|
| Venus Sentinel (ANPR) | Commercial Security System: A real-time License Plate Recognition engine deployed on Edge (RPi5) and Server. Features multi-threaded processing, WebSocket broadcasting, and a Django UI for data labeling. |
| Pill Identification System | Healthcare AI: Capable of identifying diverse medicine forms (tablets, blisters, vials) with fine-grained recognition. Includes a custom dataset workflow and API integration for mobile apps. |
| YOLO Dataset Auditor | Data Quality Tool: Automatically detects mislabeled images by extracting deep feature embeddings from YOLO models (v8-v12). Projects clusters via t-SNE and applies k-NN logic to pinpoint suspicious outliers. |
| YOLO Inference Toolkit | Open Source Tool: An OOP wrapper for Ultralytics that simplifies the developer experience. Provides a consistent "Predict → Decode" workflow for detection, segmentation, and pose estimation. |
| Deep Learning Foundations | Research to Code: Implementing seminal papers from scratch, including VGG, ResNet, and Vision Transformers (ViT), to bridge theory and practice. |
| Project | Description |
|---|---|
| Blue Archive Gacha Sim | Architecture Case Study: A feature-rich simulator built twice to compare architectures: once as a Monolithic Django app, and again as a Decoupled API (FastAPI) + SPA (Vue). |
| CZN Deck Calculator | Modern Frontend Tool: A reactive cost calculator for the game Chaos Zero Nightmare. Demonstrates complex state management and i18n support. |
