AI Data Tools

Distributed AI compute engine for scaling Python and ML workloads

Ray is an open-source distributed computing framework built for scaling AI and Python applications from a laptop to thousands of GPUs. It provides libraries for distributed training, hyperparameter tuning, model serving, reinforcement learning, and data processing under a single unified API. Ray's public site highlights OpenAI and other enterprise users. Maintained by Anyscale with Apache-2.0 open-source licensing.

Unified framework for fine-tuning 100+ large language models

LLaMA-Factory is an open-source toolkit providing a unified interface for fine-tuning over 100 LLMs and vision-language models. It supports SFT, RLHF with PPO and DPO, LoRA and QLoRA for memory-efficient training, and continuous pre-training. The LLaMA Board web UI enables no-code configuration, while CLI and YAML workflows serve advanced users. Integrates with Hugging Face, ModelScope, vLLM, and SGLang for model deployment.

Real-time search API built for AI agents

Tavily is an AI-native search API that provides real-time web search, content extraction, and crawling capabilities specifically designed for LLM applications and autonomous agents. It returns structured, citation-ready results optimized for RAG workflows with built-in safety features including prompt injection protection and PII leak prevention. Acquired by Nebius in 2026, Tavily integrates with LangChain, LlamaIndex, and major agent frameworks, serving over one million developers worldwide.

Turn websites into LLM-ready structured data

Firecrawl is a Y Combinator-backed API that crawls websites and converts them into clean, LLM-ready Markdown or structured JSON. Handles JavaScript rendering, pagination, sitemaps, and anti-bot measures automatically. Designed for RAG pipelines, AI agents, and data extraction workflows. Features batch crawling, scheduled scraping, webhook notifications, and custom extraction schemas. Processes content for direct ingestion into vector databases and LLM context windows.

2x faster LLM fine-tuning with 70% less VRAM on a single GPU

Unsloth is an open-source framework for fine-tuning large language models up to 2x faster while using 70% less VRAM. Built with custom Triton kernels, it supports 500+ model architectures including Llama 4, Qwen 3, and DeepSeek on consumer NVIDIA GPUs. Unsloth Studio adds a no-code web UI for dataset creation, training observability, model comparison, and GGUF export for Ollama and vLLM deployment.

Microsoft's open-source frontier voice AI for long-form multi-speaker audio

VibeVoice is Microsoft's open-source voice AI family with both TTS and speech recognition models. The TTS model generates up to 90 minutes of expressive multi-speaker audio with 4 distinct voices. VibeVoice-ASR transcribes 60-minute recordings in a single pass with speaker identification and timestamps. Built on continuous speech tokenizers at 7.5 Hz and next-token diffusion, it compresses audio 80x more efficiently than Encodec while preserving fidelity.

High-performance OLTP graph-vector database in Rust built on object storage for AI memory

HelixDB is an open-source, unified graph-vector database engineered in Rust that merges relational, graph, and vector workloads into a single OLTP engine, using LMDB local caching and S3 object storage for scalable agent memory.

Modular graph-based RAG pipeline using hierarchical knowledge graph community summaries

Microsoft GraphRAG is an open-source retrieval framework that transforms unstructured text into structured knowledge graphs, clusters entities hierarchically using the Leiden algorithm, and generates dataset-wide summaries alongside entity-level local search for multi-hop reasoning.

Open source business intelligence and analytics

Metabase is an open-source business intelligence and embedded analytics platform for teams that want self-service dashboards, SQL workflows, and customer-facing analytics without adopting a heavy BI suite. It supports visual querying, saved questions, alerts, database connectors, cloud or self-hosted deployment, and embedding paths that now require careful plan, permission, and license review.

ML experiment tracking and model monitoring

Weights & Biases is an AI developer platform for experiment tracking, artifact and model lineage, model monitoring, and Weave-based LLM evaluation. It helps teams log runs, compare metrics, manage datasets and model artifacts, and collaborate through dashboards, reports, alerts, SSO/RBAC controls, and hosted or self-managed deployment options.

AI-powered data quality for ML datasets

Cleanlab is a data-centric AI library that automatically detects and fixes label errors, outliers, and data quality issues in machine learning datasets. It works with any ML model and any data type including text, images, tabular, and audio by analyzing model predictions to identify mislabeled examples, near-duplicates, and ambiguous data points. Cleanlab helps teams improve model accuracy by cleaning training data rather than tuning model architecture.

Context retrieval layer for AI agents and RAG

Airweave is an open-source context retrieval platform that connects AI agents and RAG systems to 50+ apps and databases through a unified search interface. It continuously syncs data from sources like Notion, Slack, GitHub, and databases, making it searchable through LLM-friendly APIs. Airweave includes Python and TypeScript SDKs, MCP support, and a CLI for managing data connections.

Open-source toolkit for audio, music, and speech generation

Amphion is an open-source audio generation toolkit from OpenMMLab designed for reproducible research in speech synthesis, voice conversion, singing voice synthesis, and text-to-audio generation. It implements state-of-the-art models including MaskGCT, DualCodec, VITS, and VALL-E with built-in architecture visualizations for educational use. The project ships with the Emilia-Large dataset of 200,000 hours of speech data and includes multiple vocoders and evaluation metrics for benchmarking.

Open-source data curation platform for LLM fine-tuning

Argilla is an open-source platform for curating and annotating data for LLM fine-tuning and RLHF workflows. It provides collaborative annotation interfaces for text classification, ranking, and preference labeling with integrated quality metrics. Part of the Hugging Face ecosystem, Argilla supports direct dataset publishing to the Hub and integrates with major training frameworks for seamless model improvement pipelines.

Microsoft's framework for running 1-bit large language models on consumer CPUs

BitNet is Microsoft's official inference framework for 1-bit quantized large language models that enables running models with up to 100 billion parameters on standard consumer CPUs without requiring a GPU. By leveraging extreme quantization where weights use only 1.58 bits on average, BitNet achieves dramatic reductions in memory footprint and computational cost while maintaining competitive output quality for many practical use cases.

State-of-the-art open-source text-to-speech with emotion control

Chatterbox is an open-source text-to-speech model by Resemble AI that delivers state-of-the-art voice synthesis with fine-grained emotion and style control. The model supports zero-shot voice cloning from short audio samples, produces natural-sounding speech across multiple speaking styles, and runs locally without cloud dependencies. With over 24,000 GitHub stars, it has become the leading open-source alternative to commercial TTS services for developers building voice-enabled AI applications.

Real-time analytics OLAP database

ClickHouse is an open-source column-oriented database built for real-time analytical queries on massive datasets. Its columnar storage with advanced compression and vectorized query execution using SIMD instructions deliver exceptional performance for aggregations and scans. It handles billions of rows per second, supports SQL with analytical extensions, and scales horizontally for petabyte-scale data warehousing and real-time dashboards.

Knowledge graph memory engine for AI agents

Cognee is an open-source knowledge engine that builds persistent memory for AI agents by combining vector search with graph databases. It ingests data from 38+ source formats, structures information into a knowledge graph with embeddings, and enables semantic and relational queries through its ECL pipeline. Its cognitive science-inspired architecture provides superior cross-document entity identification compared to traditional RAG approaches.

Open-source deep learning text-to-speech toolkit

Coqui TTS is an open-source deep learning toolkit for text-to-speech synthesis, originally built by former Mozilla TTS engineers. It supports multi-speaker and multilingual synthesis, voice cloning from just six seconds of audio, and ships pre-trained models for 20+ languages. After Coqui shut down in 2023, the Idiap Research Institute forked and actively maintains it. With 45K+ GitHub stars, it remains the most popular open-source TTS framework in Python.

3D reconstruction without camera parameters

DUSt3R is Naver's breakthrough 3D reconstruction method that generates dense 3D scenes from unconstrained image pairs without known camera intrinsics or extrinsics. It casts pairwise reconstruction as pointmap regression, removing hard geometric constraints of projective camera models. Supports multi-view alignment, depth estimation, visual localization, and extends to MASt3R and MUSt3R for large-scale applications.

Git-based version control for ML data and pipelines

DVC (Data Version Control) is a free open-source tool that brings Git-like version control to datasets, ML models, and experiment pipelines. It stores pointer files in Git while keeping large data in remote storage like S3, GCS, or Azure. Features include reproducible ML pipelines with DAG-based dependency tracking, experiment management, metrics comparison, and a VS Code extension for visual experiment tracking.

High-performance data engine for multimodal AI workloads

Daft is a high-performance distributed data engine designed specifically for AI and multimodal workloads. It processes structured data alongside images, audio, video, and embeddings natively, outperforming Spark and Polars on AI-specific data pipelines. Built in Rust with a Python API, Daft handles the data engineering challenges unique to machine learning workflows.

Open-source BI tool for data visualization

DataEase is an open-source business intelligence tool that enables anyone to perform data analysis and build visualizations through a drag-and-drop interface without coding. It connects to MySQL, PostgreSQL, Elasticsearch, ClickHouse, and other data sources, providing interactive dashboards that can be shared via links or embedded in applications. DataEase offers chart templates, calculated fields, and role-based access control for team collaboration.

Open-source metadata platform for data discovery

DataHub is an open-source metadata platform for data discovery, governance, and observability, originally developed at LinkedIn. It provides a centralized catalog with 80+ integrations for data warehouses, lakes, dashboards, and ML platforms. DataHub offers real-time metadata ingestion, column-level lineage tracking, automated quality checks, and fine-grained access policies. Used by 3,000+ organizations in production. Apache 2.0 licensed with 11.8K+ GitHub stars.

AI data runtime for multimodal datasets and vector search

Deep Lake is an open-source AI data runtime from Activeloop for storing, versioning, and querying multimodal data and embeddings. It fits teams building RAG, training, evaluation, or dataset-heavy agent workflows that need a bridge between vector search, structured metadata, and large image, text, audio, or video collections.

Deep learning optimization for distributed training

DeepSpeed is Microsoft's open-source deep learning optimization library that makes distributed training and inference easy, efficient, and effective. Its ZeRO optimizer eliminates memory redundancies across data-parallel processes, enabling training of models with trillions of parameters. DeepSpeed supports 3D parallelism combining data, pipeline, and tensor parallelism, along with mixed precision training, gradient checkpointing, and CPU/NVMe offloading for memory-constrained environments.

ByteDance multimodal document image parser

Dolphin is ByteDance's multimodal document parsing model that handles intertwined text, tables, formulas, and figures in complex documents. Using a two-stage analyze-then-parse approach with a Swin Transformer vision encoder and MBart decoder, it performs layout analysis and parallel element parsing with heterogeneous anchor prompts. Dolphin-v2 adds document-type awareness for invoices, papers, and forms.

Lifelike AI voice generation, cloning, and voice agents

ElevenLabs is an AI voice platform for text-to-speech, voice cloning, and conversational AI agents, built on models like Multilingual v2 and the low-latency Flash v2.5 and Turbo v2.5. Developers call its API to generate lifelike narration, clone voices from short audio samples, dub content across 30+ languages, add sound effects, and deploy real-time voice agents for customer service, IVR, and interactive apps, with SDKs for Python, JavaScript, and more.

Multimodal data labeling and curation for production AI

Encord is a data labeling and curation platform for teams building production AI systems with complex multimodal data. It supports image, video, audio, DICOM medical imaging, and 3D point cloud annotation with AI-assisted labeling, advanced ontology management, and quality assurance workflows. Features active learning for prioritizing high-value samples and integrates with major ML frameworks.

Open-source feature store for machine learning

Feast is an open-source feature store that manages and serves ML features for both training and online inference. It prevents training-serving skew by providing consistent feature access across offline and real-time environments. Feast supports batch materialization from data warehouses, real-time feature retrieval, on-demand transformations, and integrates with major data platforms including BigQuery, Snowflake, Redshift, and DynamoDB.

Open-source toolkit for curating datasets and evaluating visual AI models

FiftyOne is an open-source Python toolkit from Voxel51 for building high-quality datasets and better computer-vision and multimodal AI models. It pairs a browser-based visualization App with programmatic dataset curation, embeddings, similarity search, and model-evaluation workflows.

Multilingual emotional text-to-speech with 80+ language support

Fish Speech is an open-source text-to-speech system supporting 80+ languages with emotional expression, zero-shot voice cloning, and real-time streaming. It generates natural speech with controllable emotions, speaking styles, and prosody. Features a web interface, API server, and integration with AI agent frameworks for voice-enabled applications. Over 29,000 GitHub stars.

Fast memory-efficient GPU attention kernels

FlashAttention is a fast and memory-efficient exact attention implementation that reduces GPU memory usage from quadratic to linear in sequence length. Created by Tri Dao, it achieves 3-4x speedups over baseline implementations through IO-aware tiling that minimizes HBM reads and writes. Versions include FlashAttention-2 with improved parallelism, FlashAttention-3 optimized for Hopper H100 GPUs, and FlashAttention-4 targeting Hopper and Blackwell architectures.

Open-source voice cloning and text-to-speech with few-shot learning

GPT-SoVITS is an open-source voice cloning and text-to-speech system that generates natural-sounding speech from just a few seconds of reference audio. It combines GPT-style language modeling with SoVITS voice synthesis for zero-shot and few-shot voice cloning across multiple languages. Supports Chinese, English, Japanese, Korean, and Cantonese with over 56,000 GitHub stars.

Run open-source LLMs on your phone, fully offline and private

Google AI Edge Gallery is an open-source mobile app that lets you download and run large language models like Gemma directly on Android and iOS devices with zero cloud dependency. Built on MediaPipe and LiteRT, it features AI chat with reasoning mode, multimodal image analysis, real-time audio transcription, and autonomous agent skills—all running entirely on-device for complete privacy. A reference implementation for developers building offline-first AI experiences.

Turn code and docs into a queryable knowledge graph

Graphify is an open-source AI coding assistant skill that transforms folders of code, documentation, research papers, and images into queryable knowledge graphs. It works as a skill for Claude Code, Codex, OpenCode, and other AI coding assistants. Using tree-sitter AST parsing for 19 programming languages and Claude vision for documents and images, it builds NetworkX graphs with Leiden community detection, outputting interactive HTML visualizations and structured JSON for codebase exploration.

Data quality validation framework for Python

Great Expectations is an open-source Python framework for validating, documenting, and profiling data quality. Teams define expectations as expressive unit tests for their data using an intuitive API, then validate datasets against those rules in CI/CD pipelines or production workflows. It connects to pandas, Spark, and SQL sources, generates data documentation automatically, and integrates with orchestrators like Airflow and Prefect for continuous data quality monitoring.

Synthetic data generation platform for privacy and ML

Gretel is a synthetic data platform that generates realistic, privacy-preserving datasets for ML training, testing, and data sharing. It supports tabular, text, and time-series data with configurable privacy guarantees including differential privacy. Features include data augmentation for imbalanced datasets, PII detection and anonymization, and API/SDK access for pipeline integration with BigQuery, Snowflake, and Databricks.

Agent memory system that learns, not just remembers

Hindsight is an agent memory system that enables AI agents to learn from experience rather than just store conversations. It organizes memories into three biomimetic categories: World knowledge for facts, Experiences for agent events, and Mental Models for learned understanding. The system provides retain, recall, and reflect operations backed by a temporal knowledge graph with parallel retrieval strategies including semantic, keyword, graph traversal, and temporal search.

AI Lakehouse with Feature Store for real-time ML

Hopsworks is a data-intensive AI platform combining a Python-centric Feature Store with MLOps capabilities for production ML systems. Provides sub-millisecond feature retrieval powered by RonDB, dual offline and online storage for batch and real-time inference, experiment tracking, model registry, and deployment pipelines. Available as managed cloud on AWS, Azure, and GCP, self-hosted on Kubernetes, or serverless platform.

Entity-based synthetic data generation for enterprise

K2view is an enterprise data platform that generates synthetic data using an entity-based micro-database architecture. It ensures referential integrity across complex multi-relational datasets by treating each business entity as a self-contained unit. Used for privacy-compliant test data generation, data masking, and AI training data creation in financial services, telecom, and healthcare industries.

Polyglot document intelligence framework with Rust core

Kreuzberg is a polyglot document intelligence framework with a high-performance Rust core that extracts text, metadata, images, and structured data from 91+ file formats. Available for Python, Ruby, Java, Go, PHP, C#, TypeScript, plus CLI, REST API, and MCP server. Features multiple OCR backends (Tesseract, EasyOCR, PaddleOCR), table extraction with structure preservation, and native async support.

MLSys Best Paper local RAG engine slashing vector storage by 97% on-device

LEANN is an open-source local RAG engine by StarTrail that reduces vector storage by 97% for private on-device AI assistants.

Open-source multi-type data labeling platform

Label Studio is an open-source data labeling tool by HumanSignal supporting images, text, audio, video, and time series. It offers ML-assisted pre-labeling, customizable XML-based annotation interfaces, multi-user review workflows, and REST API access. Used for computer vision, NLP, speech, and LLM fine-tuning including RLHF annotation pipelines.

Data factory for AI teams and model training

Labelbox is a comprehensive data platform for AI teams handling reinforcement learning, evaluations, robotics, and human feedback workflows. Core capabilities include RL data generation with knowledge work rubrics, custom evaluations for private benchmarks and model comparisons, robotics data with full-stack video and trajectories, and an expert network of 1.5M+ knowledge workers including 50K+ PhDs. Trusted by 80% of leading AI labs for production data operations.

Google's production on-device LLM inference framework

LiteRT-LM is Google's official open-source framework for running large language models on-device across Android, iOS, Web, Desktop, and Raspberry Pi. Already deployed in Chrome and Pixel hardware, it provides production-grade on-device LLM inference with 1.4K+ GitHub stars. Apache 2.0 licensed.

Multi-LoRA inference server for serving hundreds of fine-tuned models

LoRAX is an inference server that serves hundreds of fine-tuned LoRA models from a single base model deployment. It dynamically loads and unloads LoRA adapters on demand, sharing the base model's GPU memory across all adapters. Built on text-generation-inference with OpenAI-compatible API. Enables multi-tenant model serving without per-model GPU allocation. Over 3,700 GitHub stars.

Run and fine-tune Vision Language Models locally on Mac

Open-source Python package for running and fine-tuning Vision Language Models locally on Mac using Apple's MLX framework. Supports multimodal inference with images, audio, and video across Qwen, DeepSeek, Phi, and Gemma architectures. Features OpenAI-compatible API server, Gradio chat UI, and KV cache optimization. 3.8K+ GitHub stars.