Sanjay Nanda

Founder. AI systems architect. Engineer.

Independent thinking. Practical engineering.

Hi, I’m Sanjay Nanda N J.

Building secure AI systems and enterprise software that solve real operational problems.

Explore work Contact me
Building nowCoreframe TechnologiesFounder · Enterprise AI & software
PreviouslyBlackRockEnterprise software engineering
In the openCustosAgent governance · Public source

01 / Selected work

Substance, in the details.

All work
01
Open source · Agent governance

Custos

Public implementation

A policy boundary for AI agents.

Runtime policy evaluation and a signed audit trail for MCP tool calls. Making agent decisions inspectable, with explicit enforcement boundaries.

Runtime authorizationMCPCryptographic audit
Read the engineering notes
02
Head of Engineering · Part-time

MarginX

Engagement overview

Software for the work behind the service.

Multi-tenant restaurant software, translating operational workflows into application architecture, data models and product engineering.

Multi-tenant systemsOperational workflows
Explore the work
03
Founder & CEO · Enterprise AI & software

Coreframe

Company capabilities

Engineering within real-world constraints.

My company focuses on private AI infrastructure, enterprise knowledge systems, secure automation and cloud engineering.

Private AIEnterprise software
Explore the approach
04
Co-Founder & CTO · Creator marketplace

ORCA

Product overview

Connecting creators and brands.

Product development at the intersection of marketplace engineering, creator discovery and campaign workflows.

Marketplace systemsProduct development
Explore the product

02 / Engineering focus

From the boundary
to the whole system.

Good engineering starts with understanding what the system must do—and what it must never do.

01

Secure AI & knowledge systems

Private deployments, retrieval workflows and access boundaries. Keeping sensitive knowledge useful without making it indiscriminately available.

02

Agent infrastructure & governance

Tool authorization, policy evaluation and traceable execution. Explicit controls around what an agent can do.

03

Product & cloud engineering

Multi-tenant applications, data-intensive interfaces and maintainable services. Architecture shaped by the people operating the system.

03 / Building in public

Trust needs
something to inspect.

Custos explores a concrete question: how do you govern an agent’s tool calls and keep evidence of the decisions?

Inside Custos
Read the source Writing & ideas
Custos / Runtime governanceArchitecture
AI agentTool request
CustosPolicy evaluation
ToolAllowed execution
Signed audit recordSHA-256 chain · Ed25519 signature
Policy decisions at the tool boundary. Evidence for the calls that pass through it.

Have something in mind?

Let’s build something
worth building.

Start a conversation

Engineering challenges, thoughtful collaborations, and ideas with a purpose.