Profile banner
Subramanya Nagabhushanaradhya

Subramanya Nagabhushanaradhya

Founding ML Engineer focused on AI agent security and governance

subramanya
San Jose, California
Joined February 2025
Favicon imagesubramanya.ai
Favicon imagegithub.com
Favicon imagex.com
Favicon imagelinkedin.com
Favicon imagescholar.google.com

Network

2.7K connections
🔬
AI/ML Leaders Researchers
🎓
UMass Amherst Network
💡
Startup Founders Advisors
💰
Venture Capital Accelerators
☁️
Amazon AWS Software AI
🧠
Meta Software AI
📈
Digital Business Strategy

Summary

Focused on AI agent identity, authorization, and security: Subramanya has authored a technical specification (OIDC‑A) and multiple engineering blog posts describing how to give autonomous agents verifiable identities, delegation chains, and attestation so enterprises can safely adopt agentic workflows. arxiv+2
Builder and early‑stage founder with applied ML engineering experience: he is a founding ML engineer at Dylog and co‑founder of Agentic Trust and NockOut — combining product, security, and systems work to ship infrastructure and products for agentic/LLM systems. subramanya+2
Academic research background in multi‑modal perception and HCR: Subramanya completed an M.S. in Computer Science at UMass Amherst with research in the Human Centered Robotics Lab on vision + language for robotics / ego‑centric video understanding and maintains a Google Scholar profile. subramanya+1
Active public writer and community commentator on agentic AI safety, MCP security, and governance: maintains a regular technical blog (subramanya.ai), shares proposals publicly (arXiv / blog), and engages with the community via X/LinkedIn and conference/mentor events. subramanya+2

Work

Education

Projects

Writing

The Governance Stack: Operationalizing AI Agent Governance at Enterprise Scale

November 1, 2025

Blog/position piece proposing a five‑layer technical stack (identity/attestation, registries, policy/gateway, observability, human‑in‑the‑loop) to operationalize governance for agentic AI in enterprises.

Favicon imagesubramanya.ai

OpenID Connect for Agents (OIDC‑A) 1.0: A Standard Extension for LLM‑Based Agent Identity and Authorization

September 1, 2025

Technical proposal and specification extending OpenID Connect to represent, authenticate, attest, and authorize LLM‑based agents (defines agent claims, delegation chains, attestation formats, and flows).

Favicon imagearxiv.org

Securing MCP with OIDC & OIDC‑A: Identity‑Aware API Gateways

May 1, 2025

Blog post describing how to secure Model Context Protocol (MCP) workflows using OpenID Connect and the OIDC‑A agent extension; proposes proxy/gateway patterns for identity‑aware enforcement.

Favicon imagesubramanya.ai