Skip to content
View AnranS's full-sized avatar
🎯
Focusing
🎯
Focusing
23 followers · 79 following
  • Bytedance
  • shanghai

Achievements

Achievement: StarstruckAchievement: Pull Sharkx2Achievement: Pair ExtraordinaireAchievement: YOLO

Achievements

Achievement: StarstruckAchievement: Pull Sharkx2Achievement: Pair ExtraordinaireAchievement: YOLO

Block or report AnranS

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
AnranS/README.md

Anran — inference systems, runtimes, and developer tools. Anran’s gray-and-tan dog, with upright ears and a leaf-shaped hair clip, as a playful cartoon at a laptop.

Focus: Inference Systems Build: Runtimes and Tools Based in Shanghai

GitHub followers Stars across public repositories

👋 About Me

I'm Anran, building inference systems, runtimes, and developer tools. My projects span Rust + CUDA inference, WebAssembly tooling, game-engine integrations, and native macOS apps.

I like opening up the layers between a model and the machine—and turning what I learn into small implementations, interactive guides, and practical tools.

Understand the internals. Make the behavior visible. Build something useful.

LLM Inference · Rust & CUDA · WebAssembly · Game Tooling · Native Apps

Now building · Selected work · Contributions · Project index

🚀 Now Building & Learning

Track Projects & direction
Inference engineering nano-vllm-rs — single-GPU Qwen3 inference in Rust + CUDA, with attention validation and reproducible benchmarks.
Inference education sglang-interactive-guide — 13 Chinese chapters, browser experiments, source links, and exercises. Learn offline, without a GPU.
Game-engine tooling godot_for_minigame · tiktok-minigame-unity-demo — exports, platform SDK integration, and working API examples.
Runtimes & delivery wasm-split-tool — split large WebAssembly modules and load code on demand.
Everyday tools DesktopPulse · inline-vocab-translator — desktop widgets and tools for reading across languages.

✦ Selected Work

Project Focus What it does
nano-vllm-rs Rust CUDA Qwen3 inference on a single GPU. The inference process runs without Python, PyTorch, or libtorch; FlashAttention is optional and experimental.
SGLang Interactive Guide LLM Inference Learning An offline learning path through scheduling, KV Cache, RadixAttention, speculative decoding, and serving.
Godot Mini Game Godot SDK Integration Export Godot 4 games to WeChat and Douyin Mini Games, with TikTok Native in beta.
wasm-split-tool Rust WebAssembly Split a module into an eager primary and a lazy secondary that share memory, tables, and globals.
DesktopPulse SwiftUI WidgetKit Native macOS widgets for system health, weather, and AI-tool usage.
Inline Vocabulary Translator JavaScript Browser Tools Translate Chinese → English while reading, highlight unfamiliar vocabulary, and sync across devices.

🔬 Inside the inference work

nano-vllm-rs connects implementation to evidence:

Validation currently targets Qwen3-0.6B. Known chunked-output differences are documented in the validation report.

SGLang Interactive Guide pairs explanations with browser experiments. The simulations teach mechanisms; they are not GPU benchmarks.

🤝 Open-source Contributions

Selected merged contributions to deepcoldy/botmux, a bridge between IM platforms and AI coding CLIs:

Contribution Merged work
Codex integration PR #157 — support goal passthrough.
Resource efficiency PR #153 — reduce memory leaks, redundant disk writes, and idle screenshot overhead.
Remote configuration PR #132 — add /botconfig with owner-only configuration cards.

🛠 Tech Toolbox

Languages

Rust Python TypeScript JavaScript Swift

Inference & runtimes

CUDA WebAssembly SGLang

Apps & game tooling

SwiftUI Godot Unity GitHub Actions

🗂 Project Index

Inference & learning

Project Role Description
nano-vllm-rs Implementation Rust + CUDA inference engine and validation reports.
sglang-interactive-guide Learning guide 13 interactive chapters on inference serving.
leetcode Practice Python and Rust algorithm solutions.

Game platforms & developer tools

Project Stack Description
godot_for_minigame Godot · JavaScript Mini Game export tooling and platform integration.
tiktok-minigame-unity-demo Unity · C# Reference examples across 14 SDK API categories, with bilingual documentation.
wasm-split-tool Rust · WebAssembly On-demand loading for split WebAssembly modules.

Native apps & reading tools

Project Stack Description
DesktopPulse Swift · WidgetKit System, weather, and AI-tool usage widgets for macOS.
inline-vocab-translator JavaScript Inline translation and vocabulary learning.
wallpaper TypeScript macOS wallpaper project.
📚 Learning forks & earlier experiments

These repositories are forks or practice projects, listed separately from the work above.

Inference & ML learning forks

Runtime & tooling forks

Personal configuration & practice

📖 Documentation & releases
🏆 GitHub statistics & contribution activity
Anran's public GitHub profile statistics, dated snapshot

Snake animation of Anran's GitHub contribution grid

📫 Connect

Have a question, an edge case, or an idea for a tool? Open an issue in the relevant project—reproducible examples are always useful.

GitHub

Learning through implementation, sharing through working software.

Pinned Loading

  1. nano-vllm-rs nano-vllm-rs Public

    Single-GPU Qwen3 inference in Rust and CUDA with optional FlashAttention and reproducible benchmarks

    Rust

  2. godot_for_minigame godot_for_minigame Public

    Export Godot 4 games to WeChat and Douyin Mini Games, with TikTok Native in beta. Editor-native tooling, platform SDKs, and validated exports.

    JavaScript 57 14

  3. wasm-split-tool wasm-split-tool Public

    Split a big .wasm into an eager primary module and a lazily-loaded secondary module that share memory, table, and globals. Ship less wasm at boot, stream the rest in the background.

    Rust 4 1

  4. DesktopPulse DesktopPulse Public

    Native macOS widgets for system health, weather, and Codex, Cursor, Claude Code, and Grok usage. Built with SwiftUI and WidgetKit.

    Swift 2

  5. tiktok-minigame-unity-demo tiktok-minigame-unity-demo Public

    Reference Unity 2022.3 demo for the TikTok Mini Game SDK — covers every TT.* API across 14 categories, with bilingual code comments and README.

    C# 8

  6. inline-vocab-translator inline-vocab-translator Public

    Inline vocabulary translator for Chinese → English. Selection translate, auto-highlight unmastered words, multi-device sync via GitHub Gist or Google Drive.

    JavaScript 4 1