Enterprise AI Strategy & Agentic Governance
Define the use case, tool permissions, approval boundaries and evidence needed for an AI system. Turn those decisions into an architecture, prototype and delivery roadmap.
Discuss this workCustomize
Engineering Manager / Solution Architect – AI
I build AI systems that remember, govern, and deliver.
I design and build practical AI systems, and lead the teams that deliver them. My focus is agentic AI, runtime governance, persistent memory and cloud modernisation.
How I can help
From AI experiments to governed systems
AI Product Prototyping & Proof-of-Work Builds
Build and evaluate prototypes for agents, memory, retrieval and workflow automation. Make assumptions and limitations visible before committing to a larger investment.
Discuss this workCloud Modernisation, FinOps & Delivery Governance
Connect migration planning and engineering delivery with portfolio governance, cloud cost optimisation and AI consumption controls.
Discuss this workWorking on enterprise AI, governance, or cloud modernisation? Start a conversation
Proof of Work
Featured systems and experiments
AI Memory
llm-kosh
A local-first memory cartridge for AI assistants, giving Claude, Cursor, and MCP-compatible agents a private, persistent, searchable memory layer built on local storage and full-text search.
Agent ObservabilityAgentFabric
A telemetry and observability concept for agentic systems, exploring traces, execution metadata, decisions, tool calls, costs, and behavioural signals.
Retail AIAgentic Commerce OS
A commerce operating system concept for AI-led shopping, service, recommendations, operations, and commerce intelligence.
Agent StudioAgentKraft
A learning and creation platform for AI agents, designed around hands-on learning, intent building, studio workflows, and practical agent patterns.
AI GovernanceGovAgn
A self-hosted governance control-plane concept for autonomous AI and agent systems, focused on policy-as-code, traceability, cost governance, and audit-ready evidence.
Runtime GovernanceGavelOS
An operating-system-level governance concept for controlling agentic workflows through policy checks, enforcement layers, traceability, and system-level trust controls.
VR
Vaibhav Rastogi
From architecture to working systems
15+ years across engineering, enterprise delivery and cloud transformation, with a focus on applied AI.
I connect technical architecture with delivery: shaping use cases, building prototypes, testing assumptions and helping teams turn them into usable systems. My current work explores how AI agents retain context, act within policy, keep costs visible and leave evidence that people can inspect. The projects here include developer tools and early experiments; each is a starting point for a technical conversation.
Work with me15+Years across delivery, cloud, and product execution
6+AI systems, governance, and memory projects
Operating Thesis
The next AI layer needs memory, governance, and evidence.
The work here explores how AI-enabled organisations can keep context portable, actions governable, costs visible, and decisions traceable.
- Context and memory should be portable and inspectable
- Agent actions should be controlled near runtime
- Enterprise AI should create proof, not just promises
Principles
How I approach the work
Memory before theatreUseful AI needs inspectable context, not just larger windows and impressive demos.
Governance at runtimeAutonomous systems need policy, cost control, traceability, and audit evidence where decisions happen.
Proof over positioningThe strongest ideas become working systems that people can test, improve, and repeat.
Writing
Notes on building AI systems
Proof-of-Work Portfolio: Why Builders Need Public Evidence
A portfolio should not just describe intent. It should show working proof, clear thinking, and systems that can be tested.
Read articleRuntime Governance: The Missing Layer in Agentic AI
Agentic AI needs governance where actions happen: at runtime, across tools, policies, costs, traces, approvals, and evidence.
Read articleWhy AI Systems Need Memory, Not Just Bigger Context Windows
Bigger context windows help, but enterprise AI needs portable, inspectable, durable memory that survives sessions and supports trust.
Read articleLinks
Find me online
GitHub
Public Repositories
Python
llm-kosh
Local-first memory cartridge for Claude, Cursor, and MCP-compatible AI agents.
View repositoryPython
AgentFabric
A telemetry and observability concept for agentic systems.
View repositoryPython
AgenticCommerceOS
A commerce operating system concept for AI-led shopping, service, and operations.
View repositoryGo
Govagn
Governance control-plane concepts for autonomous AI and agent systems.
View repositoryPython
gavel-os
Runtime governance and operating-system-level control concepts for agentic workflows.
View repositoryPython
kosh
An O(1) spatial memory substrate for AI agents.
View repositoryContact
Get in touch
Have an AI governance, memory, cloud modernisation, or proof-of-work idea in mind? Send a note and I will get back to you.
Greater Reading Area, United Kingdom
Engineering Manager / Solution Architect – AI
Let’s build AI that earns trust.
Explore the projects, read the engineering notes, or get in touch to discuss AI architecture, engineering leadership and enterprise adoption.
Get in touch