最有价值项目

🖼️ Workshop: Build a multimodal AI agent with Haystack & GPT-4o — featuring image understanding, document retrieval, conversational memory, and human-in-the-loop safety controls

A comprehensive survey of Vision–Language Models: Pretrained models, fine-tuning, prompt engineering, adapters, and benchmark datasets

Universal MCP server for logistics. Connect any TMS/WMS to AI agents. Shipments, carriers, tenders, tracking — all via Model Context Protocol.

Frontend interface for building chat based system and connecting with agent driven workflows.

Semantic alignment of astronomical data with natural language using multi-modal models. (Jax) Code associated with https://arxiv.org/abs/2403.08851 (COLM 2023).

The open source implementation of the multi grouped query attention by the paper "GQA: Training Generalized Multi-Query Transformer Models from Multi-Head Checkpoints"

https://vaila.readthedocs.io/

[WACV 2025] I Dream My Painting: Connecting MLLMs and Diffusion Models via Prompt Generation for Text-Guided Multi-Mask Inpainting

WWW'24, Mirror Gradient (MG) makes multimodal recommendation models approach flat local minima easier compared to models with normal training.

Finetuning Whisper ASR model for Belarusian language

本地优先的 macOS 与 Windows 语音输入工具，支持端侧 SenseVoice 识别、全局语音输入、文件转写及可选 AI 校对。Local-first voice dictation for macOS and Windows with on-device SenseVoice, typing into any app, file transcription, and optional AI proofreading.

Built a GUI application using Tkinter that helps Doctors to prepare prescriptions more efficiently. This uses speech to text conversion and text extraction to prepare prescriptions in the correct format. This project is submitted for Smart India Hackathon 2020.

Near real-time voice processing and speech to text

Real-time offline speech-to-text transcription script on macOS using parakeet-mlx

Push-to-talk voice dictation for macOS. 100% local, free, open source. Apple Silicon MLX. No cloud, no subscription.

MLX Local Serving (MLS) - Unified ASR, TTS, and Translation on Apple Silicon

This package is the JavaScript implementation of Deepgram's WebVTT and SRT formatting. Given a transcription, this package can return a valid string to store as WebVTT or SRT caption files.

Data preparation utility for the finetuning of OpenAI's Whisper model.

Private, on-device voice assistant - STT, LLM, and TTS running fully locally. Powered by the Tether QVAC SDK. Mobile (Android via Expo + React Native) and desktop (Windows / macOS / Linux via Electron) from a shared hexagonal core.

图会议 非线性会议软件 是一个颠覆传统会议方式的协作范式，GraphMeeting 把“所有人一起开麦”变成“每人一间单人录音棚”，系统再把所有声音实时拼成一张共享思维地图。专注于通过非线性结构展现和扩充每个人的想法。它不仅仅是一个会议工具，更是一个促进讨论和协作的环境，让参与者能够自由地表达自己的观点，同时通过智能技术帮助每个想法找到可能性、可行性，并最终形成可以实践的行动路径。每个参与者的声音都会被精准捕捉，AI将自动整理讨论内容，形成一个动态、不断扩展的思维结构图，帮助团队理清复杂问题，达成共识，并推进决策和行动。