Maia

High-accuracy document parsing — PDFs and images to markdown, JSON, and HTML.

Datalab turns PDFs, images, and office documents into clean markdown, JSON, and HTML with layout, table, math, and code preservation. It is the commercial, hosted layer over the open-source Marker converter and Surya OCR toolkit, offered as a pay-as-you-go API with a free monthly allowance, while the underlying models stay free to self-host for research and small startups.

Enterprise AI platform for building, deploying, and governing ML and agentic AI.

DataRobot is an enterprise AI platform spanning the full lifecycle: predictive AI with automated machine learning (AutoML), generative and agentic AI, plus observability and governance. Its AutoML builds and compares many models at once so teams ship production-ready AI faster, and newer agent kits make enterprise AI agents practical to deploy. It runs in the cloud or on-premises for large organizations.

Full-stack document processing platform for AI agents and pipelines.

Extend is an LLM-powered document processing platform that parses, extracts, classifies, splits, and edits complex documents — handwriting, tables, and mixed formats — into reliable structured data via API or its web Studio. It combines multiple frontier models with proprietary context engineering to target 99%+ accuracy on messy real-world files. Used by teams at Brex, Square, Checkr, and Flatiron Health.

Incremental data framework for fresh AI context.

CocoIndex is an open-source data transformation framework that keeps AI agents and LLM apps supplied with continuously fresh, structured context. It turns sources like codebases, PDFs, databases, and Slack into vector or graph stores, and reprocesses only what changed (delta-only) with parallel execution by default. A Rust core drives reliability while pipelines are defined declaratively in Python, with end-to-end lineage and an observability UI called CocoInsight.

Web-scale data extraction and a knowledge graph that grounds AI in facts.

Diffbot reads the public web like a person and turns it into structured data: an Extract API for articles and products, a Crawl service, a Natural Language API, and a Knowledge Graph of billions of entities and over a trillion facts. The graph is refreshed continuously, so AI systems can ground answers in current, verifiable data rather than model memory. Diffbot also ships its own factually grounded GraphRAG language model.

AI-first intelligent document processing for end-to-end transaction automation.

Rossum reads transactional documents like invoices and purchase orders, then captures, validates, and transforms the data and pushes it into downstream ERP and approval workflows. It is built on a proprietary transactional large language model trained on tens of millions of documents that learns continuously from each customer's feedback, supports 276 languages plus handwriting, and is cloud-native. The platform targets accounts-payable and complex invoicing automation for enterprises.

Data development platform for programmatically labeling AI training data.

Enterprise platform for building and curating AI training and evaluation data with programmatic labeling instead of hand-annotating examples one by one. Teams encode domain knowledge as labeling functions that Snorkel Flow applies and refines at scale, then use the resulting datasets to fine-tune and evaluate models. Built around research from the Stanford AI Lab.

Premium human data and RLHF for frontier AI labs.

Surge AI provides high-quality human-generated training data and reinforcement learning from human feedback (RLHF) for AI developers. It pairs a large network of expert annotators with a labeling platform and API to produce complex, specialized data — code, math, safety, and domain reasoning — used to train and align frontier models. Reported customers include OpenAI, Anthropic, Google, and Meta.

Platform for building multimodal AI datasets and evaluation pipelines.

SuperAnnotate is an enterprise data platform for creating, managing, and evaluating high-quality datasets for AI. It spans annotation across images, video, text, audio, and LiDAR, with AI-assisted labeling, customizable workflows, and an optional managed annotation workforce. Teams use it to build human-in-the-loop data and evaluation pipelines for agentic, multimodal, and frontier AI.

Data factory for AI teams — labeling, evals, and human data for training.

Labelbox is a platform for generating and managing training data for AI models, combining annotation tools (Annotate), data curation (Catalog), and model-assisted labeling and evaluation (Model Foundry). It now spans reinforcement-learning data, custom evals, robotics datasets, and an on-demand network of expert human labelers, metered by a usage-based Labelbox Unit (LBU).