Trusted by regulated enterprises & agentic platforms
One question shows why this work is hard
A health insurer’s support assistant gets this question: “What’s the status of Priya Shah’s claim?” To stay safe, the company strips the name first. The model sees a placeholder and calls the claims system with it. The system finds no match, and the assistant answers “no record found.”
Nothing leaked, and the assistant is useless. Teams that hit this often switch the protection off to ship, and the risk comes back. Keeping data protected while the AI still does its job is what we work on every day.
Find sensitive data in messy text
50+ entity types plus custom ones, tested in seven languages including Arabic and Japanese.
Mask it so the model still understands
The same entity gets the same token across prompts and documents, so references stay coherent.
Control who sees the real value
Controlled unmasking by role and Context-Based Access Control (CBAC) decide who sees the original.
The code you ship here protects data for banks, health companies, and enterprise software platforms.
5,000+
live enterprise tenants at Automation Anywhere
13M+
long texts per day at a SaaS customer
99%
recall on sensitive data, including Arabic, at Bank Muscat
India AI Impact Summit 2026, New Delhi
A small team with deep experience
We are about 25 people. Our CTO, Baskaran Alagarsamy, spent over 18 years at Apple, where he led privacy engineering efforts. Our CEO, Amar Kanagaraj, started as a developer at Sun Microsystems and led product management for Microsoft Search & AI.
We take an engineering-first approach, and engineering is a mindset, not a degree. It’s the courage to tackle hard problems and the excitement of solving them simply.