platoseedplatoseed
Deep research on autopilot
Chonkie builds deep research agents that surface non-obvious signals from scattered data. Our agents combine public and private sources to deliver detailed, continuously updated research that stays current as underlying facts change. We open source most of the libraries and infrastructure we build. Our most popular library, Chonkie, handles chunking and embedding for AI applications with 100K downloads/day and adoption by OpenAI, Microsoft, and LlamaIndex.
Chonkie provides an always-on research automation platform that surfaces and summarizes signals from private and public sources, integrated with a user’s internal documents to deliver contextual, cited insights. It emphasizes proactive, topic-specific research with a UI designed for monitoring and deep-dives.
Chonkie runs ongoing research agents that monitor specified sources and surface key signals in a topic-focused UI. It can ingest private internal documents and combine them with public web data to create contextual reports, with citations for every answer. Users can ask follow-up questions or request deeper dives; reports include graphs and summarized insights without manual spreadsheet work.
Who it’s for: Organizations seeking continuous, topic-driven research and monitoring, including teams that want to combine internal documents with public data for contextual insights and proactive alerts.
Making documents AI ready at Chonkie. Previously explored on-device AI applications at Google Research and worked on leveraging organic data for better ads at Google Ads.
Feed your models better. Chonkie cleans, chunks, and makes your data AI ready
Chonkie provides an open-source data ingestion and context-building pipeline for AI projects, enabling high-quality data preparation (ingestion, cleaning, chunking, refining, and integration) to improve accuracy, speed, and cost efficiency. It offers both open-source installations and hosted/ On-prem options for builders and AI-native businesses.
Monitor AI agents and understand user behavior
Database of every product on the internet
