ENGINEERING · RESEARCH · ADVISORY

Security for
what comes
next.

I’m Sagar. I secure cloud and AI systems at Greenlight, build open-source tools, and help founders turn hard security problems into useful products.

Step into my workThe story so far
A MAP OF MY PRACTICE

01 / BUILD & OPERATE

Cloud, identity, data, and AI security. Built around the realities of the teams using it.

Inside my engineering work
THE THROUGH LINE

Powerful systems deserve thoughtful boundaries.

01 / THE WORKBENCH

Questions worth asking.
Tools worth building.

Open-source tools and research exploring what autonomous agents should be allowed to do.

01AGENT SECURITYPython package · Beta

Agent Circuit
BreakerACB

Before an agent acts,
ask whether it should.

A checkpoint between an AI agent’s proposed action and its execution. Deterministic checks help trusted integrations identify risky commands, require approval, and keep a record of what happened.

Creator & maintainer
Explore the project Source code
Listed on FAUN.toolbox
AT THE DECISION BOUNDARY

What happens when an agent tries this?

PROPOSED COMMANDgit push --force origin main
BLOCK

Decision: do not execute

A force push can overwrite shared history. ACB flags this destructive command shape.

Documented examples. No commands run.Try it yourself
02AGENT CONTAINMENTResearch prototype

KECHAAKernel-Enforced Containment
for Hijacked AI Agents

What if the agent
stops following the plan?

Research into keeping agent tools within an approved task, even when instructions are hijacked. Signed permissions and an experimental Linux backend explore how to enforce those limits beyond the prompt.

Developer / researcher
Explore the research
PERMISSION HAS A PERIMETER
Conceptual illustrationExperimental Linux backend
ANOTHER LINE OF INQUIRYHorizon Agent ContainmentHow does risk build across a sequence of agent actions?

02 / THE PERSON BEHIND THE WORK

BENGALURU, INDIA
Sagar Chhatrala

Sagar ChhatralaEngineer. Researcher. Always building.

A builder’s instinct.
A researcher’s curiosity.

I care about the moment an idea becomes something a team can actually use.

At Greenlight, I lead cloud and AI security, helping protect financial infrastructure and data serving millions of families. My independent work follows the same question into AI agents: how do we give powerful systems useful limits?

I also advise Rig Security on identity security and partner with AirMDR on product design, bringing the perspective of someone who builds and operates these systems.

01 / STAY CLOSE

Start with the real problem.

Build around the people who will use the system and the decisions they need to make.

02 / MAKE IT EXPLICIT

Give trust a boundary.

Be clear about what a system can do, where permission ends, and what remains uncertain.

03 / KEEP IT OPEN

Share what you learn.

Turn practice into code, research, standards, and conversations others can build on.

THE LONG VIEWThe work behind
the perspective.Explore my experience

03 / OUT IN THE OPEN

From the workbench
to the wider room.

Sharing practical perspectives on cloud, data, identity, and AI security.

PUTTING PRACTICE INTO STANDARDS

CIS benchmark co-author

AWS Foundations v1.5.0
Microsoft Azure Foundations v2.0.0

2026CrowdStrike Fal.ConCloud, data & AI security2025CrowdStrike Fal.ConRemoving Greenlight’s Blind SpotsSPEAKERAI Builders NetworkConference speaker

Also at Cisco SecCon & OWASP Bangalore.

The full speaking story

04 / AN OPEN INVITATION

Have a hard
problem?
I’m listening.

A product taking shape. A room full of curious people. A question about where AI security goes next. Let’s compare notes.

LinkedIn GitHub
TO: SAGAR

What brings you here?

Draft an email
OR, STRAIGHT TO THE SOURCEchhatralasagar99@gmail.com