Blog
/
Authors
/
Shubham Jindal Dir

Author Profile

Shubham Jindal Dir

AI Software Engineering

at
Harness

Shubham Jindal is Director of AI at Harness, where he leads the company's AI strategy and the unified AI platform that powers products across the organization. An engineer at heart and a systems thinker by nature, he focuses on turning cutting-edge AI into reliable, governed, and scalable enterprise outcomes, bridging the gap between experimentation and adoption to help teams build AI.

Connect on LinkedIn

Latest from

Shubham Jindal Dir

A pinned write-up worth starting with.

Featured

Harness AI

Identity and Permissions for AI Worker Agents in Harness

How Harness scopes AI agent access with delegated identity, ephemeral tokens, RBAC, OPA policy, and MCP gateway tool intersection, enforced server-side.

Shivansh Srivastava
Shubham Jindal Dir
+ more
July 16, 2026
18 minutes to read
Read the article
Identity and Permissions for AI Worker Agents in Harness

All posts by

Shubham Jindal Dir

All
Engineering Blog
Technical
Operations

How We Build

Engineering

How We Secured AI Worker Agents in Harness

How Harness isolates Autonomous Worker Agents with four independent security layers — image hardening, process isolation, secret isolation, and network controls, tested against a real CVSS-9.0 breach chain.

July 13, 2026
14 minutes to read
How We Secured AI Worker Agents in Harness

Engineering

Identity and Permissions for AI Worker Agents in Harness

How Harness scopes AI agent access with delegated identity, ephemeral tokens, RBAC, OPA policy, and MCP gateway tool intersection, enforced server-side.

July 16, 2026
18 minutes to read
Identity and Permissions for AI Worker Agents in Harness

Engineering

Architecting MCP for AI Agents: Lessons from Our Redesign

The Harness MCP v2 Server covers 125+ resource types across 30 toolsets with 11 tools and 26 prompt templates. It supports Cursor, Claude Code, and any MCP-compatible client.

March 19, 2026
13 minutes to read
Architecting MCP for AI Agents: Lessons from Our Redesign

How We Build

Engineering

How We Secured AI Worker Agents in Harness

How Harness isolates Autonomous Worker Agents with four independent security layers — image hardening, process isolation, secret isolation, and network controls, tested against a real CVSS-9.0 breach chain.

July 13, 2026
14 minutes to read
How We Secured AI Worker Agents in Harness

Engineering

Identity and Permissions for AI Worker Agents in Harness

How Harness scopes AI agent access with delegated identity, ephemeral tokens, RBAC, OPA policy, and MCP gateway tool intersection, enforced server-side.

July 16, 2026
18 minutes to read
Identity and Permissions for AI Worker Agents in Harness

Engineering

AI Demos Are Easy. Enterprise AI Is Not.

Why Context is the New Code for Scaling Production AI

April 6, 2026
9 minutes to read
AI Demos Are Easy. Enterprise AI Is Not.

Engineering

Architecting MCP for AI Agents: Lessons from Our Redesign

The Harness MCP v2 Server covers 125+ resource types across 30 toolsets with 11 tools and 26 prompt templates. It supports Cursor, Claude Code, and any MCP-compatible client.

March 19, 2026
13 minutes to read
Architecting MCP for AI Agents: Lessons from Our Redesign
No items found.
No items found.

More voices from the Harness team

Engineers, security leads, and platform folks worth following.

All authors
Tom O'Reilly
Tom O'Reilly

Senior Vice President, Worldwide GTM Partnerships

Peter Hack
Peter Hack

Sales Engineer, Global Alliances