-blue?style=for-the-badge){kind=icon}
"I don't just write code — I fight for every byte of RAM."
I'm a high school student who fell in love with the space between hardware and software — the place where a single misplaced volatile keyword can crash a whole system, and where a well-placed DMA transfer can make a $3 microcontroller feel like magic.
Everything here is self-taught: no bootcamp, no formal curriculum — just datasheets, forum threads at 2 AM, and a lot of bricked boards along the way.
|
🔩 Hardware-level optimization OPI PSRAM tuning, bare-metal programming when Arduino abstractions get in the way |
🧵 Real concurrency FreeRTOS dual-core task management, mutexes, race-condition hunting on ESP32-S3 |
🤖 Edge AI/ML On-device inference and on-device learning — no cloud dependency |
🎓 Next chapter: aiming to continue my studies at PENS (Politeknik Elektronika Negeri Surabaya) to turn this self-taught foundation into formal, rigorous expertise.
🌱 Currently exploring (click to expand)
- 🔬 Deeper bare-metal ESP-IDF programming (peeling back the Arduino layer)
- 🧠 On-device TinyML model compression
- 🖥️ Custom bootloader & OTA reliability for the Accretion Phone
- 📡 Low-power mesh networking patterns
Like a planet forming from dust and debris colliding over millions of years, every project below accreted onto the skills of the last. Nothing here was built in isolation — each device is a fossil record of what I learned building the one before it.
graph LR
A[🕹️ G-BOX<br/>2D Game Loop] --> B[📡 AI-Pocket<br/>IoT + Power]
B --> C[🎬 Media Hub<br/>Color + Web]
C --> D[📷 SANZXCAM<br/>Vision + Sensors]
D --> E[🎧 RYNE Engine<br/>Edge ML]
E --> F[📱 Accretion Phone<br/>The OS]
style A fill:#1a1a2e,stroke:#a855f7,color:#fff
style B fill:#1a1a2e,stroke:#a855f7,color:#fff
style C fill:#1a1a2e,stroke:#a855f7,color:#fff
style D fill:#1a1a2e,stroke:#a855f7,color:#fff
style E fill:#1a1a2e,stroke:#a855f7,color:#fff
style F fill:#1a1a2e,stroke:#ffd700,color:#fff
| Platform | ESP32 · OLED 128×64 SSD1306 (I2C) |
| Games shipped | Flappy Bird · Space Invaders · Asteroids |
My first custom handheld console — and the place where it all began. Pure fundamentals: building a real game loop from scratch, pushing pixels manually to a monochrome buffer, wrestling with array-based memory management, and relearning trigonometry (sin/cos) not for a math test, but to make 2D physics actually feel right.
💡 Lesson learned: if you can render a bird flapping through pipes on a 128×64 screen with no game engine, you can render almost anything.
| Platform | ESP32-C3 SuperMini · OLED 128×64 |
My first real dive into the internet and power efficiency:
- 🔋 Battery Guardian — physical voltage-divider circuit + firmware to catch battery drain/leakage in real time
- ⌨️ Custom 6-button virtual keyboard for full text input, wired to Google Gemini 2.0 Flash
- 📻 Internet-free radio messaging built on ESP-NOW Mesh Networking
💡 Lesson learned: you don't need Wi-Fi to build a network — you need a protocol and a reason to talk to your neighbors.
| Platform | ESP32-S3 N16R8 · TFT ST7789 (320×170) |
Leveling up to color displays and real web data processing:
- 🎞️
.mjpegvideo playback via DMA, powered by LovyanGFX & JPEGDEC - 🛰️ Real-time APIs: ISS Tracker, NASA APOD, Groq AI (parsed with
ArduinoJson) - 🧩 2D game AI using BFS pathfinding for enemy movement in Pacman & Tank Battle
💡 Lesson learned: BFS isn't just a whiteboard interview question — it's what makes a Pacman ghost feel like it's actually hunting you.
| Platform | ESP32-S3 · 16-bit DVP Camera · MPU6050 · ILI9341 |
A deep dive into image processing and system stability:
- 🌀 Kalman Filtering sensor fusion for real-time Electronic Image Stabilization
- 🌤️ HDR Triple Capture for wider dynamic range
- 🏝️ Dynamic Island–style UI with sprite-based restoration for tear-free animation
- 🧠 On-device analysis via Gemini 2.5 Flash Lite (Mood Reader, ANPR)
- 🔐 Working steganography — payloads hidden in a JPEG's COM marker & a BMP's LSB bits
💡 Lesson learned: stabilizing a handheld viewfinder with a $2 gyroscope taught me more about sensor fusion than any tutorial ever could.
| Platform | ESP32-WROVER · OLED SSD1306 · Bluetooth A2DP Source |
The big leap into local Edge Machine Learning:
- 🧮 Softmax Multiclass Vibe Classifier — a self-learning model in PSRAM that retrains every 5 minutes, inferring mood from skip-rate & volume
- 🎲 RYNE-CRS, a custom recommendation algorithm (not UCB1) built on Bayesian probability + a power-law curiosity model
- 😃 Expressive OLED face animations, synced with a FreeRTOS Core 1 task watching audio buffer health for glitch-free playback
💡 Lesson learned: an AI model doesn't need a data center — it needs a well-scoped problem and a PSRAM chip that isn't afraid of a little math.
| Platform | ESP32-S3 · ILI9341 (Resistive Touch) · MPU6050 · SD Card (SDIO 1-bit) |
| Codebase | 13,000+ lines · zero real OS underneath |
| Apps built-in | 26 |
The culmination of every project above:
- 🔒 Lock screen with genuine glassmorphism
- 🏠 Home screen with momentum scrolling (low-pass velocity + decay), flicker-free thanks to aggressive double buffering / canvas sprite optimization in PSRAM
- 🤖 Gemini AI integration — voice dictation via INMP441 mic + hand-written manual JSON parser (no
ArduinoJson) to save RAM - 📡 OTA updates over Wi-Fi or SD card
- 🎮 Game Mode: "Inferno", a 3D raycasting FPS à la DOOM/Wolfenstein, stable 16 FPS, persistent autosave to SD
💡 This isn't just a phone. It's every lesson from G-BOX to RYNE Engine, compressed into a device that fits in your pocket.
|
Languages
|
Frameworks & Libraries
|
Hardware / Chips
|
Tools & Services
|
"Mastery isn't a single breakthrough — it's accretion. It's ten thousand small, unglamorous hours of debugging, rewiring, and rewriting, quietly fusing together until one day the dust becomes a planet."
— SanzX