Skip to content
You must be logged in to sponsor uno-km

Become a sponsor to EUNHO KIM

@uno-km

EUNHO KIM

uno-km
Seongnam, South Korea

💖 Sponsor @uno-km & AMEVA On-Device AI

Hi, I am uno-km (@uno-km), an indie systems hacker building the AMEVA On-Device AI Ecosystem.

"Even without Big Tech's multi-million dollar H100 clusters, I wanted to prove that true on-device AI can be conquered right inside our pockets—using everyday phones and vintage silicon."


😭 A Bitter, Tearful Confession: Why Sponsor Me?

No corporate fluff. Here is the unvarnished reality of my daily lab:

  1. The Tragic & Comical Arsenal in My Pocket:
    • My testbed: a Galaxy A35, an old Galaxy S21, and a Galaxy S25 bought on a 24-month loan just to benchmark Snapdragon 8 Elite.
    • A friend said: "My Galaxy S20 screen is shattered, I'm tossing it." I fell to my knees: "DON'T THROW IT AWAY! I can run Whisper STT on that!" I rescued it, and I run tests while poking fingers through cracked glass.
    • My mom shouted: "Where did my old Galaxy S7 in the drawer go?!" ...Mom, forgive me. Your 2016 phone is compiling Android Bionic libc kernels 24/7 in my room.
  2. Surviving on Instant Ramen & Airport Wi-Fi:
    • I live on instant ramen and save every penny for cheap store coffee to stay awake coding through the dawn.
    • My room's Wi-Fi drops every 10 minutes. When pushing huge GGUF models or npm/pip packages, I pack my laptop and cracked phones, rush to Incheon Airport or sit outside a 24-hour store freezing in the cold, piggybacking on public Wi-Fi to hit git push origin main. People think I'm flying abroad; nope, I'm just shipping code.

Yet through all this misery, I never gave up.


🔥 What We Conquered: Pure Mobile Silicon Madness

Big Tech told the world: "Local AI on a phone is impossible. Pay our $50/month API fee."

We refused. We couldn't bear to see mobile GPUs sitting idle, so we broke open mobile Vulkan SPIR-V compute shaders on Android Termux, and rewrote 14 native libraries with Android Bionic libc and ARM64 NEON!

0% Root Required! $0 Cloud Bills! 100% Offline Edge Autonomy!

  • ⚡ Termux-BitNet: World’s first mobile native port of MS 1.58-bit ternary LLM! Zero CMake compile hell—1-line packaging with NEON + Vulkan GPU in 10s.
  • 🎙️ Termux-TTS: Sub-100ms 22.05kHz on-device speech synthesis. Lightweight DSP to Vulkan GPU studio neural nets with a 4-Tier self-healing fallback.
  • 👁️ Termux-Vision: Eradicated 150MB+ OpenCV bloat. Pure Canny/Sobel + SmolVLM Vulkan GPU zero-copy visual intelligence on phone cameras.
  • 🔗 Termux-AIChain: Zero dependencies—50KB ultra-lightweight autonomous agent orchestrator running on pure topological DAG graphs.
  • 🦙 Termux-LlamaCpp: Ended cooking your phone for 20 minutes with Clang. Pre-compiled Bionic binaries running GGUF LLMs in 10 seconds.
  • 🛡️ STT · Diffusion · Train · Playwright · Forge · Sentinel · Runtime · MCP-Hub · Infra-Index: Offline Whisper STT, non-root Samsung Gallery image generation, on-device LoRA fine-tuning, and browser WebGPU tensor engines.

💰 Where Your Money Actually Goes

No corporate BS. No fake plaques. Your money goes directly into survival and silicon:

  1. Rescuing Used Test Phones (Top Priority): Buying secondhand Snapdragon, Exynos (Xclipse AMD GPU), Dimensity, and Tensor devices to guarantee zero crashes on every Android phone on earth.
  2. Hacker Sustenance: Coffee to keep me awake, warm soup, and a reliable internet line so I don't freeze at the airport.
  3. 24/7 Phone Rack Maintenance: Cooling fans and power hubs so phones dangling from power strips don't catch fire during 10,000-iteration stress tests.

🤝 Realistic Sponsorship Tiers

  • ☕ $5/mo — Coffee Backer: Buys coffee that keeps me coding through the dawn. Name etched in SPONSORS.md.
  • 🍜 $25/mo — Meal & Cable Backer: Provides a hot meal and fast cables. Prioritized issue answers on GitHub.
  • 📱 $100/mo — Adopt-a-Used-Phone Hero: Funds 1 used Android phone for our test rack! Sponsor badge on all 14 flagship READMEs + benchmark logs.
  • 🚀 $500/mo — Core Patron: Keeps the solo developer coding full-time. Logo & link on official portals + 1:1 C++/Vulkan advisory.
  • 👑 $2,000+/mo — Enterprise Partner: Custom kernel profiling for your target hardware, partner banner on portal & release notes, and deep architectural sessions.

💬 "Code Feedback & Brutal Critiques Are Always Welcomed"

We are the "Grassroots Stream"—resource-constrained hackers squeezing maximum performance out of humble hardware against multi-billion dollar cloud monopolies.

If you spot a flaw in my code, open an Issue or PR. I will take the critique, fix it, and prove our worth with lighter and faster code.


"Your cup of coffee turns a smartphone with a shattered screen into a supercomputer."

Featured work

  1. uno-km/termux-playwright

    "The definitive, production-grade Playwright & Chromium automation toolkit for Android Termux. Zero root, anti-bot stealth, and crash-resilient."

    Python 3
  2. uno-km/termux-diffusion

    Production On-Device AI Image Generation Framework for Android Termux & Samsung Galaxy. Dual-Engine (Python & Node.js), Bionic ARM DotProd/FP16 SIMD, Vulkan GPU acceleration.

    Python 2
  3. uno-km/termux-stt

    🎙️ Production-grade On-Device Speech-to-Text & 128d X-Vector Speaker Diarization for Android Termux. whisper.cpp, Vosk, Sherpa-ONNX, and 0 PyTorch Dependency.

    Python 1
  4. uno-km/termux-bitnet

    ⚡ Production 1.58-bit (i2_s) BitNet On-Device Inference SDK & Dual Engine for Android Termux & ARM64.

    C++ 2

Select a tier

Choose a custom amount.

$5 a month

Select

Listed in official repository SPONSORS.md directory and AOSF Discord supporter badge.

$25 a month

Select

Profile link featured on the official Foundation website and priority issue response.

$100 a month

Select

Silver sponsor badge displayed across all 6 flagship repository READMEs and architecture Q&A advisory.

$500 a month

Select

Corporate logo and permanent link featured on main site, PyPI/npm headers, and customized edge AI PoC advisory.

$2,000 a month

Select

Advisory Board seat, joint press releases, and dedicated NPU/GPU kernel optimization workshops.