AYR

Princeton-based IDP provider developing patent-pending synthetic training data generation to accelerate document processing deployments from months to days.

AYR

Overview

Founded in 2018 as Singularity Systems, AYR built its market position around a single, persistent IDP friction point: organizations cannot share sensitive real documents for model training, and document formats change constantly, forcing repeated retraining cycles. The company's answer is synthetic data generation — creating structurally diverse training documents without exposing regulated or proprietary records.

That approach earned AYR Everest Group Leader status in 2023 after three consecutive years as Major Contender, and a Top 30 Fastest Growing Company recognition the same year. Enterprise validation followed through BNY Mellon's Accelerator Program, signaling a customer pipeline in financial services.

By February 2026, analyst firm Deep Analysis included AYR in its Vendor Vignettes subscriber series alongside Salesforce, Celonis, UiPath, and Xceptor — labeling it "earlier stage" relative to those platform peers but recommending it for shortlisting in complex financial document processing projects. The placement is a signal of enterprise evaluation, not validated dominance: no benchmark figures or head-to-head comparisons are publicly available from that assessment.

Dr. Tianhao Wu, CTO, frames AYR's competitive position directly against legacy players: "What takes other legacy OCR and classic IDP companies upwards of nine months, a team of data scientists, and thousands of samples to only achieve 60%-80% accuracy, AYR can achieve as high as 99.9% in 48 hours." That claim is unverified by independent benchmarks; buyers should request test data directly.

How AYR Processes Documents

AYR's core architecture combines its patent-pending AI Pathfinder Technology — a multi-modal approach using Computer Vision, NLP, and proprietary OCR engines — with the Intelligent Document Simulator (IDS), which generates synthetic training data to accelerate model readiness.

The February 2026 IDS upgrade materially extended what the tool can do. Where the prior version could only substitute random field values drawn from user-provided dictionaries, the updated release adds structural manipulation: users input a single sample document and generate a wide range of layout variants by swapping table columns and shuffling document sections horizontally or vertically. As Dr. Wu described it: "This makes it possible to train AI and machine learning models to recognize and process a wide range of document layouts, which is crucial for handling the diverse and ever-changing documents that businesses deal with daily."

Capability	Previous IDS	Current IDS (Feb 2026)
Field variation	Random values from user-provided dictionaries	Retained
Layout manipulation	Not available	Swap table columns, shuffle sections horizontally/vertically
Training document diversity	Limited to structural copies of one sample	Wide layout range from a single sample document

The practical effect is that AI models can train on structurally diverse documents without any access to live business data — compressing the data preparation phase that typically consumes weeks or months of an IDP deployment. Availability status (GA, beta, or preview) and pricing for the updated IDS were not disclosed at launch.

The underlying SingularityAI Platform combines the IDS with the Data Perfection Platform (DPP) for user customization and AI model engineering, and claims proprietary OCR superior to leading commercial alternatives. No independent OCR benchmark substantiates that claim publicly.

Use Cases

Financial Services with Data Confidentiality Requirements

Organizations processing financial documents face a structural problem: sharing real training data violates confidentiality requirements. AYR's synthetic data generation sidesteps this by creating document variations — including structural layout variants — without exposing sensitive customer information. Deep Analysis specifically recommends AYR for shortlisting in projects involving complex, multi-page invoices and other financial documents, though the recommended use case is narrow enough to suggest AYR is not yet positioned as a general-purpose IDP platform.

A telecommunications VP reported: "I gave them our hardest, most complex unstructured invoices. Within a week, AYR was able to give me an intelligent AI model with an API... and was able to get 95% accuracy on the first runs."

Rapid Deployment for Complex Document Processing

Customer case studies show 96% improved data entry accuracy, 50% manual labor reduction, and $4M+ annual cost savings for commercial loan processing, with 13x ROI increases achieved through AYR's accelerated deployment approach. These figures are sourced from AYR's own materials; no third-party audit has been published.

Technical Specifications

Feature	Specification
Training Data Generation	Patent-pending synthetic document creation with field variation and structural layout manipulation (column swapping, section shuffling)
Input Required	Single sample document sufficient to generate wide layout range
Model Training Time	Hours to days (vs. industry standard of months; unverified by independent benchmark)
Claimed Accuracy	99.9% in 48 hours (vendor claim; no independent verification published)
Core Technology	Multi-modal AI (Computer Vision, NLP), proprietary OCR, LLM integration (proprietary + GPT-3)
Platform	SingularityAI Platform with Intelligent Document Simulator and Data Perfection Platform
Pricing / Availability	Not publicly disclosed

Resources

Website
Intelligent Document Processing Solutions
SingularityAI Platform
Blog
Deep Analysis Vendor Vignette (paywalled; public summary available)

For buyers evaluating AYR for financial document processing: pricing, plan tiers, IDS availability status, and independent accuracy benchmarks are not publicly available and should be requested directly from the vendor.

AYR