The secure, verifiable path to enterprise AI - built for the places the public cloud can't go.
Most AI demonstrations collapse the moment a lawyer, a regulator, or a defense officer asks a single question: where did that answer come from? Pryon built its entire product around that question. Founded in 2017 in Raleigh, North Carolina, the company makes a secure retrieval-augmented generation platform - RAG, in the industry's shorthand - that ingests an organization's own documents, audio, images and video, then pairs that content with large language models to return fast answers that carry a citation for every line.
The founder's résumé explains the ambition. Igor Jablokov previously built Yap, an early cloud voice-recognition company that became Amazon's first AI-related acquisition; its inventions helped seed Alexa, Echo and Fire TV. Before that, as a program director at IBM, he led the team that designed the precursor to Watson and built the world's first multimodal web browser. Pryon, according to industry lore, was even the internal code name for the speech engine that eventually became Alexa.
That history matters because Pryon is not chasing consumer novelty. It sells to enterprises and government agencies in sectors where a wrong or unattributed answer carries real cost: defense, energy, life sciences, manufacturing and finance. The pitch is deliberately unglamorous - proven, safe and trusted - and aimed squarely at buyers who have watched the generative-AI hype cycle and grown wary of tools that hallucinate.
The technical bet sits underneath the chatbot everyone else is racing to add. Pryon spent years on content ingestion: connecting to enterprise repositories, then normalizing, deduplicating, chunking and vectorizing messy multimodal content into a retrievable knowledge base. The company describes this as a Memory Layer - and in 2025 it won recognition as AI Data Management Solution of the Year in the 8th annual AI Breakthrough Awards, a category it shared with names like NVIDIA, Microsoft and Databricks despite a team of roughly 130 people.
Deployment is where Pryon draws its sharpest line against rivals. The RAG Suite is built to run air-gapped, on-premises, in a private cloud, in a federal cloud, in a multi-tenant public cloud, or in hybrid combinations - all from one product. For a Department of Defense entity or a regulated utility, that flexibility is not a convenience; it is the entire reason a deal can happen at all.
Ingest · Retrieve · Attribute · Deploy anywhere
An end-to-end retrieval-augmented generation platform. It ingests text, audio, images and video, then pairs it with LLMs to deliver source-attributed answers - reportedly across hundreds of thousands of pages per collection and thousands of concurrent users, with millisecond responses.
Automated connectors pull from enterprise repositories, then normalize, deduplicate and federate multimodal content into a retrievable knowledge base - the unglamorous plumbing that makes accurate retrieval possible.
A data-management layer that persists and governs organizational knowledge for retrieval. Named AI Data Management Solution of the Year in the 2025 AI Breakthrough Awards.
“We intentionally built a product for enterprises that is secure and reliable - qualities many now realize are lacking in the generative AI dominating the news cycle.”— Igor Jablokov, Founder & Chairman, Pryon
Enterprises already own the answers their teams need - the information is just buried across incompatible systems, formats and file types. Pryon's thesis is that the product isn't new knowledge; it's faster, verifiable access to what an organization already has, without letting sensitive content leave the perimeter.
Deployment flexibility — one product, every environment
Source: Pryon product documentation. Bars indicate supported deployment modes, not performance benchmarks.
Horizontal enterprise-search tools and cloud-native RAG stacks compete for the same attention, and rivals like Glean, Cohere and Vectara chase adjacent problems. Pryon differentiates on three things regulated buyers care about most.
Every answer is traceable back to the content it came from, with human-feedback loops - designed for settings where an ungrounded response is unacceptable.
Sensitive content never has to touch the public internet. Enterprise-grade security and air-gapped operation are first-class features, not add-ons.
Built to answer in milliseconds across very large content collections and thousands of concurrent users, without trading away accuracy.
Built Yap (Amazon's first AI acquisition, which seeded Alexa) and led the IBM team behind the precursor to Watson. Founded Pryon in 2017. Awarded Eisenhower and Truman National Security fellowships.
Named President and CEO in 2024. Brings two decades of enterprise-software experience, including formative roles at early Salesforce and Informatica, focused on scaling go-to-market.
Subscription and licensing of the RAG platform to large enterprises and government agencies, plus integration services - concentrated in regulated, high-security verticals.
Igor Jablokov starts the company to bring secure, accurate AI to enterprise knowledge - years before generative AI hit the mainstream.
Pryon develops its content-ingestion and retrieval platform and raises early venture capital from Aperture, Good Growth and Triangle-area funds.
Closes a round led by Thomas Tull's US Innovative Technology Fund to scale indexing and analysis of enterprise data.
Chris Mahl becomes President and CEO; founder Igor Jablokov transitions to Chairman.
Pryon's AI Memory Layer is recognized in the 8th annual AI Breakthrough Awards.
Pryon sits in the enterprise RAG and knowledge-AI market, competing with tools like Glean, Cohere and Vectara and the RAG stacks of Microsoft Azure AI, AWS and NVIDIA. Its wedge is regulated, high-security buyers who value verifiability and control over consumer-style polish. Reported customers and partners span the spectrum:
Pryon builds a secure enterprise retrieval-augmented generation (RAG) platform that ingests an organization's content - text, audio, images and video - and pairs it with large language models to deliver fast, accurate, source-attributed answers.
Pryon was founded in 2017 by Igor Jablokov, the entrepreneur behind Yap (Amazon's first AI acquisition, which seeded Alexa). Chris Mahl became President and CEO in 2024, with Jablokov serving as Chairman.
Pryon has raised more than $140M in total, including a $100M Series B in September 2023 led by Thomas Tull's US Innovative Technology Fund.
Large enterprises and government/defense agencies in regulated sectors, with reported customers and partners including NVIDIA, Dell Technologies, the World Economic Forum, Acrisure, Westinghouse and U.S. Department of Defense entities.
Pryon emphasizes verifiable, source-attributed answers, enterprise-grade security, and flexible deployment - including fully air-gapped and on-premises environments - rather than consumer-facing novelty.