Headline

Co-Founder at AgentLabs

About

I am building AgentLabs, an enterprise software company for a world where people and AI agents work inside the same governed systems. The premise is easy to state and hard to build. Agents are becoming capable enough to act inside real business systems rather than draft text beside them. The moment an agent can post to a ledger, approve a leave request, or call a tool against production data, four questions need answers: who is acting, what are they allowed to do, who approves the consequential steps, and what actually happened. Most enterprise AI today answers none of them. So we are building the layer that does. AgentForge runs agents as untrusted workloads inside enforced isolation, where capabilities are granted by intersection and a missing grant fails closed. AgentLabs Auth treats people, services, and agents as principals, so an agent's authority is delegated and scoped rather than inherited from someone's login. On top of those foundations, we build the business systems ourselves, from planning and delivery to people operations and accounting, because a platform that has never run real work is a claim. My work runs across all of it: the Go runtime and control plane, the Linux isolation model, identity and authorisation design, the product surfaces, and the delivery method our own human-agent teams work by. Reasoning may be probabilistic. Authority must not be. That sentence is most of what I think about. Based in Bengaluru. If you are working on agent identity, enterprise governance, or the systems agents will act inside, I am glad to talk.

Location

Bengaluru

Work experience

Co-Founder at AgentLabs Technologies Pvt ltd, Aug 2026 - Aug 2026 · 1 mo, Bangalore, Building the platform and products for governed human-agent work: AgentForge for agent identity, isolated execution and evidence; AgentLabs Auth for identity and authorisation across people, services and agents; AgentLabs Notify for attributable communication. Business systems on the same foundations: Agentbooks, Agentlabs HRMS and AgentSpace. I work on the parts that have to be right: the execution boundary, the identity model, and the record of what happened. The runtime runs agents as untrusted workloads inside a Linux sandbox built on namespaces, cgroup v2, seccomp and Landlock, where an agent's capabilities are the intersection of what it declared, what policy allows, and what its grant permits, and anything missing fails closed. Identity is an OpenID Connect provider issuing tokens to human, delegated-agent and workload principals, with each service enforcing policy deny-by-default. Tenancy is enforced by row-level security in the database, not in application code. Around that: versioned Protobuf contracts, immutable content-addressed workload images, and durable sessions backed by a write-ahead log.

Director of Engineering at AppSentinels, Apr 2023 - Aug 2026 · 3 yrs 5 mos, Bengaluru

Engineering Leader at Twilio, May 2022 - Apr 2023 · 1 yr

Engineering Leader at Cisco, Nov 2020 - Dec 2022 · 2 yrs 2 mos, Greater Bengaluru Area

Executive Director at NGX Technologies Pvt Ltd, Sep 2007 - Nov 2020 · 13 yrs 3 mos, Bengaluru Area, India

Co-Founder at NGX Technologies Pvt Ltd, Jan 2007 - Nov 2020 · 13 yrs 11 mos

Education

Bachelor of Engineering (BE), Electrical, Electronics and Communications Engineering, Visvesvaraya Technological University