A customer uploads a contract, a scanned form or a spreadsheet. Before an AI system can use it, the application needs to read it correctly. [1][4]
Tables, handwriting and unusual layouts make that first step harder. An answer built on a missing row can be wrong before the model starts reasoning. [4]
Reducto turns those documents into structured data.
Its platform parses files, extracts requested fields, splits document bundles and fills completed forms. Citations tie the output back to the original page. [1][3]
For the AI team, document understanding becomes a service the application can build on. [3]
The document tasks behind the application
Parse reads layouts, text, tables and figures, returning document structure with bounding-box citations. Extract turns the fields a team asks for into schema-based output. [1]
Split separates bundles into individual documents. Edit fills blanks, tables and checkboxes using supplied data. Those tasks can become stages in the same document pipeline. [1]
Developers can use the API and SDKs. Studio gives teams a way to build document pipelines without writing code, while CLI and MCP support bring document tools into agent workflows. [3]
Reducto also supports hosted, VPC, on-premises and air-gapped deployments for different enterprise requirements. [3]
A document platform built by an MIT team
Raunak Chowdhuri and Adit Abraham founded Reducto in 2023. Both studied computer science at MIT; their backgrounds include computer vision research, machine learning and search products. [2]
The company joined Y Combinator’s Winter 2024 batch. Its initial document ingestion work has grown into a platform covering more of the tasks around a document. [2][3]
What this looks like inside customer products
Harvey: making difficult legal documents readable
Harvey’s early ingestion tools handled many documents, but complex scans, handwriting and redlines exposed gaps. The legal AI company evaluated Reducto against real customer edge cases. [4]
Harvey moved from evaluation to production in roughly six weeks, including an on-premises Kubernetes deployment. Built-in citations let the application trace content back to the document. [4]
After rollout, Harvey’s team saw fewer OCR-related customer concerns and greater use of OCR on difficult documents. [4]
Vanta: returning the completed questionnaire
Vanta uses Reducto to parse security questionnaires and supporting documents. Vanta’s AI then prepares answers using that information. [5]
The remaining job was returning a filled document in its original format. Vanta and Reducto developed document filling together, moving from a prototype to production within a month. [5]
The completed flow runs from an uploaded questionnaire to answers written back into the file. Vanta also uses Reducto in compliance-document evaluation. [5]
From reading the file to completing the work
Reducto’s r-1 parsing model is available in preview. It handles layout and document structure together as part of the wider platform. [1]
The company’s direction extends beyond turning a PDF into text. Parsing, extraction and editing let an application carry a document task through to a finished artifact. [3][5]
Teams can try their documents in Studio before integrating the platform into production. [1][3]
Where Reducto ranks
Current rankings. Positions come from public evidence, and no company can pay for a place. How we rank
The bigger picture
Legal and compliance applications depend on the details inside their customers’ files. [4][5]
Reducto handles the reading, structure and document operations underneath those applications. [3]
Turn the document into data, then put that data to work.
Written by SOTA2 from Reducto’s public materials. Customer experiences and product plans are reported by Reducto. Ranking analysis and comparison data are from SOTA2.