Hey there! Thanks for stopping by my little corner of the internet. I'm 張小凡(whyb), and I'm thrilled to have you here. 🎉

I'm a passionate engineer deeply immersed in the world of visual image processing, AI deployment, and next-generation developer workflows. Lately, I've shifted my lens towards the speech domain, exploring the intricate architectures of ASR and TTS and OCR models.

Beyond core AI, I am a hardcore practitioner and enthusiast of the modern "Vibe Coding" movement. I have a profound obsession with building AI Agent workflows, crafting Model Context Protocol (MCP) servers, and designing automated Agent Skills to seamlessly bridge the gap between LLMs and real-world software control.

• 🔭 Working at a company building cutting-edge tools for audio processing and video editing.
• 👁️‍🗨️ Exploring computer vision applications—from image understanding to intelligent visual effects.
• 🤖 Building robust MCP servers and autonomous agent skills to supercharge local development and application automation.

I'm proud to be a part of some amazing open-source projects, including:

• HwCodecDetect is a lightweight CLI/GUI tool for inspecting hardware-accelerated video codecs, supporting capability matrix checks across different resolutions, bit depths (8/10/12-bit), and chroma formats (YUV 420/422/444).
• Gomoku-AI is implemented with PyTorch, designed with reference to AlphaZero. It takes Player 1 as the main training target, while Player 2 serves as a sparring model to help Player 1 improve. Leveraging deep learning, it simulates players' move strategies through the training mechanism of a neural network model (based on the deep structure of ResNet and reinforcement learning). [Web Game]
• PinyinIME is a Windows TSF (Text Services Framework) pinyin input method built in C++17 around a three-tier architecture: a lightweight DLL text service injected into applications, a single-instance dictionary server communicating through sandbox-compatible shared-memory IPC, and a Win32 settings/registration UI, featuring a pre-compiled binary dictionary, DP pinyin segmentation, fast prefix completion, user dictionary learning, fuzzy pinyin, and simplified/traditional conversion.
• cifa.js is a WebAssembly-powered Web Online IDE & Script Playground for the Cifa script engine (a lightweight, elegant C-style interpreter originally by @scarsty). It brings the mini scripting engine to the browser via Emscripten/Wasm, featuring a Monaco Editor with syntax highlighting, real-time Linting, sandboxed execution inside Web Workers, and a VS-style error console. [Live Playground]
• FastChwHwcConverter is a high-performance, CPU multi-threaded(OpenMP/OneTBB), GPU acceleration(ROCm/CUDA), header-only C++ library for converting image data formats between HWC (Height, Width, Channels) and CHW (Channels, Height, Width).

• Tencent's ncnn is a high-performance neural network inference framework. Pure C++ implementation, cross-platform(Windows, MacOS, Linux, Android, iOS), and supports cpu(loongarch, arm, risc-v, mips, x86), supports GPU(Nvidia, AMD, Intel, Qualcomm, Apple, Hisilicon, Rockchip npu, Axera, etc..)
• inisis's OnnxSlim is a lightweight Python toolkit designed to optimize ONNX models.
• nihui's ruapu is using various extended instructions to harass and amuse the CPU to inferred whether the CPU supports a certain extended instruction set.

• 💻 Languages: Modern C++ – my trusty sidekicks.
• 🌍 Location: Based in the vibrant city of Chengdu.

C++ OpenCV OpenVINO Onnxruntime DirectML TensorRT ffmpeg OpenGL Vulkan Qt

Feel free to explore my repositories, and don't hesitate to reach out if you want to collaborate or just chat about tech. Happy coding! 😄