🌏 China | 🤖 Building AI companions | 🚀 Full-stack agentic engineer

Building the future of human-AI interaction — from terminal agents to desktop companions with voice, vision, and presence.

• 🤖 Ani — AI-powered desktop companion with immersive scenes, real-time video interaction, and natural voice conversation
• 🧠 Codev — Collaborative Agentic Workspace combining search, browser control, coding, voice, and long-session continuity
• 🧠 Omni — Multimodal LLM training & inference framework (LM, VLM, speech) from scratch, with real-time voice/video call
• 📚 CS336 — Hands-on implementation of Stanford CS336: LLM training, tokenization, transformers, and scaling
• 🔄 Duplex — Real-time bidirectional translation with voice I/O and unified UI

• 🤖 Ani — AI-powered desktop companion with wallpaper integration, video layers, companion mode, and plugin system
• 🎙️ Voice Engine — STT/TTS pipeline with VAD, noise suppression, and multi-provider support (Groq, faster-whisper, edge-tts)
• 🧩 Plugin Runtime — Modular plugin system with card protocol, UI panels, and SDK discovery

• 🧠 Codev — Full-featured agentic coding workspace with ToolSearch, SubAgent Swarm, and auto-compact context management
• 🖼️ ImageShowTool — Terminal-native image rendering with Kitty protocol, ink-picture, and aspect-ratio-aware display
• 🔍 WebSearch + WebFetch — Multi-backend search (SearXNG, Tavily, Jina) with image search and terminal image results
• 🗺️ LocationTool — Geographic search with Amap & Google Maps, unlimited paginated results
• 🔌 Bridge System — CCR remote forwarding, AskUserQuestion decoupling, DingTalk & Feishu integration

• 🧠 Omni — Multimodal LLM framework: train LM, VLM, and full speech models from scratch; real-time voice/video call
• 🔄 Duplex — Real-time bidirectional translation with voice I/O, Groq Whisper ASR, and unified Qt UI
• 📚 CS336 — Transformer implementations from scratch, GPT-2 training, scaling laws
• 🔧 System Tooling — Custom IPC, auto-update pipelines, data epoch migration, cross-session collaboration

• Building Ani — Creating an AI companion that feels truly present: immersive wallpaper scenes, real-time voice, and natural interaction
• Expanding Codev — Making agentic development workflows faster, smoother, and more collaborative
• Building Omni — A from-scratch multimodal LLM framework spanning text, vision, and speech
• Studying LLMs — Working through Stanford CS336 to deepen understanding of transformer internals
• Rapid prototyping — Full features in days, iterating on real user feedback

"Build what you'd want to use yourself" — Every project starts as a personal need, then grows into something others can benefit from.