About Taos
We're building the governance layer for enterprise AI agents.
Taos makes AI agents safe to deploy in regulated industries — by enforcing authorization, audit, and rollback at the moment of execution, before anything happens.
Our story
Enterprise AI adoption is running ahead of enterprise AI governance. Agent frameworks like LangGraph, CrewAI, and AutoGen make it straightforward to build capable agents. None of them answer the questions that matter most in regulated industries: who authorized this action, can it be undone, what data did the model see, and where is the audit trail? Those aren't questions you can bolt on after the fact.
Taos was built by a team with deep experience in the compliance regimes that govern financial services, healthcare, and government — SOC 2, SOX, PCI-DSS, HIPAA, FedRAMP. We've been on the inside of audits, incident responses, and the engineering work required to satisfy them. The architecture reflects what we know auditors actually need, not what sounds good in a security whitepaper.
The result is a Rust execution kernel that runs inside your network — between every AI agent and every system it calls. It enforces policy before tool calls are dispatched, writes tamper-evident audit records before operations proceed, and provides a cryptographic delegation chain from human to agent to tool for every action taken. It's the infrastructure layer that makes AI agents auditable by design, not by accident.
Team
Akon Dey
Founder
30+ years across startups and big companies in the US, Australia, and India. Ran Dgraph Labs as CEO — cloud graph database through growth and acquisition. At Visa, led architecture for commercial solutions: application performance, security, compliance, governance, and oversight for the commercial card business. Earlier: Yahoo!, Awake Security (Arista), Platfora (Workday), focused on distributed databases, streaming, and analytics at scale. PhD on transactions across heterogeneous data stores; still focused on transactional correctness, OLTP and OLAP, and highly available systems.
Abhishake Gajja
Founder
30+ years across financial services and fintech, healthcare, logistics, and insurance — sectors where a weak ML rollout becomes a regulatory incident, not a harmless experiment. Built ML and AI from greenfield into production-grade deployments: owned models, monitored pipelines, and the narrative when risk and compliance push back hard. Knows the gap between notebook prototypes and production sign-off — drift, lineage, controls, and what auditors actually ask for. IIT Bombay roots in analytics and BI; Python and large-scale data stacks; turns ambiguous stakeholder asks into systems people actually run.
Patrick Farry
Founder
20+ years designing secure, scalable platforms for logistics, fintech, and IoT — from edge AI orchestration on constrained devices to event-driven microservices supporting 700,000+ connected devices. Led security programs to PCI-DSS Level 1 and ISO 27001 compliance. Specialises in enterprise architecture, Zero Trust, and engineering governance.