Cognaize: IDP Software Vendor
On This Page
New York-based intelligent document processing company specializing in neuro-symbolic AI for financial services, backed by $18M Series A funding.
Overview
Founded in 2020, Cognaize is led by CEO Al Eisaian - a repeat founder with a prior exit to VMware - and CTO and CPO Vahe Andonians, whose prior fintech company was acquired by Moody's. The company raised $18 million in Series A funding led by Argonautic Ventures (Viken Douzdjian, Managing Partner) with participation from Metaplanet (Rauno Miljand, Managing Partner), reporting 4x ARR growth in 2022. Valuation was confirmed in the "hundreds of millions" but not disclosed precisely.
The core strategic bet is domain specificity over general-purpose AI. Eisaian frames the market as three options for financial institutions: build in-house (impractical), deploy general AI via consultants (expensive and imprecise), or use Cognaize. The platform is trained on 1.3 million financial documents - not a general-purpose LLM fine-tuned on financial text - covering loan applications, SEC filings, ESG disclosures, presentations, and trustee reports. Customers include two of the three largest credit ratings agencies and large insurers, all unnamed.
Human-in-the-loop is a product architecture decision, not a compliance footnote: financial analysts correct model outputs and act on results within the designed workflow. Andonians emphasizes data privacy - "We will continue to refine our client-side AI models, guaranteeing nobody gains an advantage of client's data except the client itself" - addressing auditability requirements in regulated financial environments.
The company added Knowledge Graph integration in May 2023, enabling semantic search and cross-document intelligence, and won the 2025 A-Team Group Innovation Award for Knowledge Graph Excellence. Membership as an extraordinary FinTech member of the Association of German Banks since September 2024 reflects active European expansion, supported by offices in Germany and Armenia alongside the New York headquarters. Whether the domain-specific model advantage holds as general-purpose models improve on financial tasks is the open question the Series A does not answer.
How Cognaize Processes Documents
Cognaize combines neural networks for pattern recognition with symbolic AI for logical reasoning and financial domain knowledge - an architecture the company calls neuro-symbolic AI. The neural layer handles unstructured document variability; the symbolic layer applies financial rules and relationships that pure deep learning models must infer from data alone.
Documents enter the platform and are matched against 30+ financial templates or processed through the domain-trained model. The Knowledge Graph layer enables semantic search across documents and cross-document intelligence extraction - connecting entities and relationships that span multiple filings or agreements rather than treating each document in isolation.
Human-in-the-loop validation is embedded throughout: analysts review model outputs, correct extractions, and make decisions based on results. This workflow design makes the system auditable by construction, which matters for credit agencies and insurers operating under regulatory scrutiny. Client-side AI model deployment means document data is processed locally, so no client data is exposed to third parties or used to train shared models.
Use Cases
Credit and Lending
Processes 120,000 documents annually with 99.7% accuracy, reducing manual spreading time by 75% and cutting costs 10x for major banks analyzing complex loan documents and credit agreements. The platform handles variable document structures across credit agreements without requiring per-template configuration.
Ratings and Research
Two of the three largest credit ratings agencies use Cognaize for financial report validation. 500,000 reports processed across 12 million pages achieving 99.9% accuracy with a 66% reduction in processing time. The human-in-the-loop architecture supports the auditability requirements ratings agencies face when justifying analytical outputs to regulators.
ESG and Compliance Disclosure
Financial institutions extract structured data from ESG-related documents using hybrid intelligence workflows, achieving 99.3% classification accuracy while reducing manual processing by 80%. The Knowledge Graph integration enables cross-document ESG analysis - connecting disclosures across reporting periods and entities.
Technical Specifications
| Feature | Specification |
|---|---|
| AI Architecture | Neuro-symbolic (neural networks + symbolic AI) |
| Training Data | 1.3M+ financial documents |
| Document Types | Credit agreements, SEC filings, ESG disclosures, loan applications, trustee reports, presentations |
| Accuracy | 99.9% (reports), 99.7% (credit agreements), 99.3% (classification) |
| Templates | 30+ financial templates |
| Language Support | Multi-language |
| Deployment | Client-side AI models, on-premises, cloud, hybrid |
| Compliance | GDPR, financial regulations |
| Performance | 75-80% time reduction, 10x cost savings |
| Knowledge Graph | Semantic search, cross-document intelligence extraction |
| Human-in-the-Loop | Embedded analyst validation workflows |
Resources
Company Information
Headquarters: New York, United States (115 Broadway); additional offices in Germany and Armenia
Founded: 2020
CEO: Al Eisaian (prior exit to VMware)
CTO and CPO: Vahe Andonians (prior fintech acquired by Moody's)
Funding: $18M Series A - Argonautic Ventures (lead), Metaplanet (participant); valuation in "hundreds of millions," undisclosed