Projects

### Overview: Este repositório é o conteúdo em português para a AWS AI Practitioner (AIF-C01), focando em fundamentos de IA/ML na AWS, serviços gerenciados (Bedrock, Sagemaker, etc.), boas práticas de governança e aplicações de negócio. O repositório é organizado em módulos objetivos, com foco em leitura progressiva, revisão rápida e prática guiada, para quem está iniciando no escopo do exame AIF-C01. ### Key features: - **Fundamentos de IA/ML na AWS**: Conceitos de IA, ML, inferência, LLMs e base da certificação. - **Serviços de IA AWS**: Bedrock, SageMaker, serviços especializados e integração com aplicações. - **Governança e Segurança**: Governança, conformidade, privacidade, revisão humana e controles essenciais. - **Revisão**: Simulados curtos, questões comentadas, glossário e material de fixação. - **Apoio**: Cheatsheets, flashcards, labs guiados e recursos complementares. ### Tech stack: HTML ### Why it matters: Este repositório é o conteúdo em português para a AWS AI Practitioner (AIF-C01), focando em fundamentos de IA/ML na AWS, serviços gerenciados (Bedrock, Sagemaker, etc.), boas práticas de governança e aplicações

Overview: deployKF is a platform that combines Kubeflow, Airflow, and MLflow into a complete machine learning platform for Kubernetes, providing easy deployment and maintenance for any Kubernetes user. Key features: - Run on any Kubernetes cluster, including on-premises and in the cloud - Intuitive centralized configurations for all aspects of the platform - Seamless in-place upgrades and config updates - Connects to existing tools Tech stack: Shell Why it matters: deployKF offers a managed service experience with the flexibility of a self-hosted solution, making it accessible for any Kubernetes user.

Overview: This repository contains Java backend interview questions and project实战s, covering topics such as Java basics, JVM, concurrent programming, Spring, Spring Boot, Spring Cloud, MySQL, Redis, message queues, distributed systems, microservices, and AI agents. It also includes detailed answers and a study route. Key features: - Java backend interview questions - Project实战s (13 projects that can be included in resumes) - AI tutorials (Spring AI, RAG, Agent) - Detailed answers and study route Tech stack: Java, Spring, Spring Boot, Spring Cloud, MySQL, Redis, message queues, distributed systems, microservices, AI agents Why it matters: This repository provides a comprehensive resource for Java backend developers, offering practical experience and detailed answers to help them prepare for interviews and improve their skills. Repository: dvsusan/susanSayJava GitHub stars: 462 GitHub description: Java 面试八股文 + 企业级项目实战：Spring、Spring Boot、Spring Cloud、JVM、Redis、MySQL、分布式、微服务、AI Agent，附详细答案与学习路线。 README excerpt: # Java 面试八股文与企业级项目实战（持续更新） Java 后端面试题与项目实战合集：覆盖 Java 基础、JVM、并发编程、Spring、Spring Boot、Spring Cloud、MySQL、Redis、消息队列、分库分表、高并发、微服务、分布式与 AI Agent，每篇都是可落地的实战经验。 -

### Overview: This Node.js package provides a fast and production-ready question answering solution using the powerful DistilBERT model, leveraging the Hugging Face Tokenizers library for efficient text processing and TensorFlow.js for model inference. ### Key features: - **Production-Ready**: Automatically downloads and uses the default DistilBERT model in SavedModel format. - **3 Lines of Code**: Minimal code required to integrate question answering directly into Node.js applications. - **Model Compatibility**: Compatible with various DistilBERT, BERT, and RoBERTa models in both SavedModel and TFJS formats. - **Remote Models**: Supports remote models via TensorFlow Serving for enhanced scalability. ### Tech stack: Node.js, TypeScript, Hugging Face Tokenizers, TensorFlow.js. ### Why it matters: This package simplifies the process of integrating question answering capabilities into Node.js applications, making it an ideal choice for developers looking to enhance their applications with intelligent text analysis.

Overview: HuggingSound is a toolkit designed for speech-related tasks using Hugging Face's tools, aiming to provide a simple and user-friendly environment for experiments. Key features: - Speech recognition using CTC models hosted on Hugging Face Hub - Inference capabilities with a language model boost - Evaluation metrics for speech recognition tasks Tech stack: Python Why it matters: This toolkit simplifies speech-related experiments, making it accessible for developers and researchers without the need for complex setup or extensive knowledge of speech processing libraries. Repository: jonatasgrosman/huggingsound GitHub stars: 463

Overview: YOLOFace is a project that utilizes the YOLOv3 algorithm for real-time face detection, leveraging the power of deep learning. Key features: - Real-time face detection using YOLOv3 algorithm - Supports image, video, and webcam inputs - Pre-trained YOLOv3 weights file for face detection Tech stack: Python Why it matters: This project showcases the capabilities of YOLOv3 in real-time face detection, offering a practical application of deep learning techniques in computer vision.

Overview: UrbanGPT is a spatio-temporal large language model that was accepted by KDD'2024. It integrates a spatio-temporal dependency encoder with instruction-tuning to enhance generalization capabilities. Key features: - Seamless integration of spatio-temporal dependency encoder and instruction-tuning paradigm - Extensive experimental findings highlighting potential for spatio-temporal learning in zero-shot scenarios - Models and data available on Huggingface for further details Tech stack: Python Why it matters: UrbanGPT demonstrates the potential of large language models for spatio-temporal learning, especially in zero-shot scenarios, showcasing exceptional generalization capabilities across various urban tasks.

### Overview: Whisper-GUI is a simple GUI application designed to facilitate the use of Whisper, a powerful text-to-speech and speech-to-text model. This tool simplifies the process of transcribing audio and video files, automatically or manually detecting the language, and providing timestamps for both words and sentences. It supports multiple output formats, including SRT, JSON, and TXT, and is compatible with various models from Whisper and WhisperX. ### Key Features: - Transcribe local audio and video files. - Automatic or manual language detection. - Word and sentence-level timestamps. - Multiple output formats: SRT, JSON, TXT. - Supports multiple State-of-the-art Whisper & WhisperX models. - Advanced and customizable transcription options and optimizations. ### Tech Stack: Python ### Why It Matters: Whisper-GUI simplifies the process of transcribing audio and video files, making it easier for developers and anyone interested in text-to-speech and speech-to-text technology to use Whisper without the need for complex command-line interfaces.

### Overview: RLAIF-V is an open-source framework designed to enhance the trustworthiness of large language models (LLMs) through the utilization of open-source feedback. By leveraging high-quality feedback data and an online feedback learning algorithm, RLAIF-V achieves super GPT-4V-level trustworthiness across both generative and discriminative tasks. ### Key Features: - **Super GPT-4V Trustworthiness via Open-source Feedback**: RLAIF-V 12B achieves super GPT-4V trustworthiness in both generative and discriminative tasks by learning from open-source AI feedback. - **High-quality Generalizable Feedback Data**: The feedback data used by RLAIF-V effectively reduces the hallucination of different MLLMs. - **Inference-time Scaling by RLAIF-V reward**: RLAIF-V reward consistently improves the trustworthiness performance of different models when scaling up inference budgets. ### Tech Stack: Python ### Why It Matters: By providing a framework that maximally exploits open-source feedback, RLAIF-V ensures the trustworthiness of large language models, making them more reliable and accountable in various applications.

### Overview: JARVIS-ChatGPT is a conversational assistant powered by synthetic voices, including J.A.R.V.I.S. from IronMan, equipped with OpenAI Whisper, ChatGPT, and IBM Watson APIs. It generates meaningful answers with a badass voice and can interface with PCs, servers, and external gadgets. ### Key Features: - **Synthetic Voices:** J.A.R.V.I.S. voice from IronMan - **OpenAI Whisper:** For natural language processing - **ChatGPT:** For generating human-like responses - **IBM Watson:** For additional capabilities - **Tacotron Model:** For voice generation ### Tech Stack: Python ### Why it Matters: JARVIS-ChatGPT is designed to be a versatile and powerful assistant that can provide real-time tips and opinions on any topic, accessible through any authorized microphone. It can interface with PCs, servers, and external gadgets, making it a valuable tool for researchers and decision-makers. The project aims to streamline research processes and provide informed insights, all while ensuring stability and efficiency.

### Overview: A 3D vision library designed for monocular and stereo 3D detection of humans, social distancing, and body orientation. Monoloco is built on three research projects and is continuously tested on Linux, macOS, and Windows. ### Key features: - Monocular 3D Human Localization - Stereo 3D Human Localization - Social Distancing Detection - Body Orientation Estimation ### Tech stack: Python ### Why it matters: Monoloco is a powerful tool for researchers and developers interested in 3D human localization, social distancing, and body orientation detection. Its robust performance and continuous testing make it an essential resource in the field of computer vision and robotics.

Overview: Adam-mini is a PyTorch implementation of Adam optimizer that significantly reduces memory usage while maintaining or improving performance compared to AdamW. Key features: - Adam-mini uses fewer learning rates (lr) in the Adam algorithm. - It partitions parameters into blocks and assigns a single but good lr to each block. - The resulting algorithm is shown in Algorithm 1. - Supports popular distributed frameworks including DDP, FSDP, DeepSpeed, and Torchtitan. Tech stack: Python Why it matters: Adam-mini reduces memory usage by 50% compared to AdamW while achieving comparable or better performance, making it a valuable tool for models that require efficient memory management. Repository: zyushun/Adam-mini GitHub stars: 462

Overview: This repository contains a curated list of popular AI infrastructures designed for machine learning training and inference in production environments. It covers architectures, inference at scale, compiler and optimization stacks, and distributed training platforms. Key features: - End-to-end machine learning training pipelines - Inference at scale on Cloud or end devices - Compiler and optimization stacks for deployment on various devices - Novel ideas for efficient large-scale distributed training Tech stack: Python, TensorFlow, Kubernetes, Docker, Git, Markdown, and GitHub. Why it matters: This list aims to provide a comprehensive overview of AI solutions in production, focusing on overall architectures rather than individual frameworks. It serves as a learning resource for understanding how to build robust systems that enable machine learning in production. Repository: 1duo/awesome-ai-infrastructures GitHub stars: 461 GitHub description: Infrastructures™ for Machine Learning Training/Inference in Production. README excerpt: A curated list of AI infrastructures for machine/deep learning training and inference in production. It includes TensorFlow Extended, Kubeflow, and other platforms. The goal is to provide a comprehensive overview of AI solutions in production, focusing on overall architectures rather than individual frameworks. The list is for learning purposes and welcomes contributions.

### Overview: This repository serves as a curated list of resources aimed at conserving, developing, and documenting low-resource languages. It includes open-source code and tools useful for various aspects of language documentation, such as lexicography, annotation, text automation, and more. ### Key Features: - **Documentation and Lexicography**: Resources for creating dictionaries and lexicographic materials. - **Software Tools**: Various software utilities for language documentation and development. - **Annotation**: Tools and resources for annotating language data. - **Format Specifications**: Guidelines and specifications for formatting language documents. - **i18n-related Repositories**: Repositories focused on internationalization and localization. - **Audio Automation**: Tools for automating audio processing. - **Text-to-Speech (TTS)**: Systems for converting text to speech. - **Automatic Speech Recognition (ASR)**: Systems for automatic speech recognition. - **Text Automation**: Tools for automating language text processing. - **Experimentation**: Resources for conducting experiments related to language documentation. - **Flashcards**: Tools for creating and managing language flashcards. - **Natural Language Generation**: Systems for generating natural language text. - **Computing Systems**: Resources for developing and maintaining computing systems. - **Android Applications**: Applications for Android devices. - **Chrome Extensions**: Extensions for Chrome browsers. - **FieldDB**: Webservices, components, and plugins for FieldDB. - **Academic Research Paper-Specific Repositories**: Repositories dedicated to academic research papers. - **Example Repositories**: Examples of repositories that can be used as references

**Overview:** Visual-Chinese-LLaMA-Alpaca（VisualCLA）是一个基于中文LLaMA&Alpaca大模型的多模态中文大模型，旨在增强模型对视觉信息的理解和处理能力。通过在中文LLaMA/Alpaca模型上添加图像编码模块，VisualCLA能够接收和理解视觉信息，并通过多模态预训练和指令精调进一步提升其多模态理解和执行能力。 **Key features:** - 支持图像和文本输入的中文多模态模型 - 多模态预训练和指令精调 - 提供推理代码和基于Gradio/Text-Generation-WebUI的部署脚本 - 多模态指令理解任务效果展示 - 开放翻译测试集 **Tech stack:** Python **Why it matters:** VisualCLA通过增强中文LLaMA/Alpaca模型的多模态处理能力，为跨模态信息的整合和应用提供了新的解决方案，特别是在自然语言处理和多模态任务中。

### Overview: These examples showcase Jina in action and provide sample code for you to work from. They cover a range of applications, including semantic search, cross-modal search, and advanced text and image retrieval. The examples are designed to help beginners understand how Jina works and to provide a foundation for more advanced users. ### Key Features: - Semantic Wikipedia Search with Transformers and DistilBERT - Search Lyrics with Transformers and PyTorch - Find Similar Audio Clips - Querying While Indexing in the Wikipedia Search Example - Cross Modal: Search images from captions and vice-versa - App Store Search - Star Wars Question Answering System - Financial Question Answering Search ### Tech Stack: Python ### Why It Matters: These examples demonstrate the versatility and power of Jina in various domains, from text search to image retrieval, and provide a solid foundation for developers to build their own applications.

Overview: MS-G3D is a PyTorch implementation of "Disentangling and Unifying Graph Convolutions for Skeleton-Based Action Recognition," presented at CVPR 2020. This project focuses on improving action recognition through graph convolutional neural networks, specifically designed for skeleton-based actions. Key features: - PyTorch implementation - Supports NTU RGB+D 60, NTU RGB+D 120, and Kinetics 400 datasets - Includes data preprocessing and skeleton generation tools - Pretrained models available for quick results on NTU RGB+D 60, NTU RGB+D 120, and Kinetics 400 Tech stack: Python, PyTorch, NVIDIA Apex, PyYAML, tqdm, tensorboardX Why it matters: This project offers a significant advancement in action recognition by leveraging graph convolutional networks, making it a valuable tool for researchers and developers in the field of computer vision and robotics. Repository: kenziyuliu/MS-G3D Primary language: Python GitHub stars: 462 GitHub description: [CVPR 2020 Oral] PyTorch implementation of "Disentangling and Unifying Graph Convolutions for Skeleton-Based Action Recognition", CVPR 2020 Oral.

### Overview: This OpenFrameworks addon integrates Google's TensorFlow library, providing a comprehensive set of tools for building and manipulating machine learning models within the framework of OpenFrameworks. ### Key Features: - **TensorFlow r1.1 Support**: The addon supports the latest TensorFlow version r1.1. - **OpenFrameworks 0.9.8 Compatibility**: It is compatible with the latest version of OpenFrameworks. - **Precompiled Libraries**: Available for Linux (GPU and CPU) and macOS (CPU-only), with no support for Windows. - **Example Code**: Includes examples for loading and manipulating different types of TensorFlow models in OpenFrameworks, such as image-to-image, style transfer, and sequence generation. ### Tech Stack: - **C++**: The backend is written in C++. - **Python Bindings**: The library has Python bindings for easier integration. ### Why It Matters: This addon simplifies the process of integrating TensorFlow models into OpenFrameworks projects, making it easier for developers to leverage advanced machine learning techniques within their projects.

### Overview: PremSQL is an open-source library designed to help developers create secure, fully local Text-to-SQL solutions using small language models. It provides all the essential tools to build and deploy end-to-end Text-to-SQL pipelines with customizable components, making it ideal for secure, autonomous AI-powered data analysis. ### Key features: - **Local-First**: Avoid third-party closed-source providers and keep your data secure. - **Multiple connectors**: Supports PremAI, Ollama, HuggingFace, Apple MLX, and OpenAI. - **Customizable Datasets**: Create, fine-tune, and evaluate models with built-in or custom datasets. - **Robust Executors and Evaluators**: Easily connect to databases and assess model performance. - **Advanced Generators**: Convert natural language prompts into executable SQL queries. - **Error Handling and Self-Correction**: Automatically correct SQL queries during inference. - **Fine-Tuning Support**: Fine-tune models with LoRA, QLoRA, or full fine-tuning strategies. - **Agents**: Use PremSQL baseline agent to perform Text to SQL, write analysis reports, and plot simple charts on databases. - **Playground**: Use our playground to do the same for agents but with a better ChatGPT UI like experience dedicated for AI powered data analysis. - **Importing CSVs or Kaggle CSV dataset directly to PremSQL playground**: You can analyse any CSV dataset from kaggle directly or from any folder using PremSQL. ### Tech stack: Python ### Why it matters: PremSQL is an essential tool for

Overview: This repository is a curated collection of resources, papers, tools, and implementations that bridge the gap between Retrieval-Augmented Generation (RAG) and Reasoning in Large Language Models and Agents. It aims to enable more powerful Agentic AI systems by bringing together traditionally separate research domains. Key features: - Comprehensive list of RAG and Reasoning resources - Papers and papers related to RAG and Reasoning - Tools and implementations for RAG and Reasoning - Industry-leading deep research products for Agentic Deep Research Tech stack: Python, JavaScript, and other relevant technologies. Why it matters: This repository serves as a valuable resource for researchers and practitioners interested in combining the strengths of RAG and Reasoning to enhance the performance of large language models and agents. By bridging the traditionally separate domains, it provides a comprehensive collection of resources and tools for researchers to explore and contribute to the field of Agentic AI. Repository: DavidZWZ/Awesome-RAG-Reasoning GitHub stars: 460 GitHub description: [EMNLP 2025] Awesome RAG Reasoning Resources