Discovery
Customer conversations, workflow mapping, product judgment and finding the real operational bottleneck behind a vague AI request.
I build the unexpected into business workflows
Open to AI systems, agent infrastructure and forward deployed engineering.
What I Do
From ambiguity to production
I usually enter before there is a clean technical specification. I work with customers to understand the workflow, identify where AI creates leverage, design the system, build it and get it running in production.
Architecture
Agents, memory, tools, multimodal systems, RAG, approvals, integrations and data models that can survive real usage.
Deployment
Full-stack engineering, APIs, background workers, infrastructure, observability and the iteration loop after launch.
iKawn Work
Agentic innovations that show my thinking
These are the systems where the product idea, AI architecture, interface, integrations and production constraints meet in one place.
iKawn OpenBrain
A persistent intelligence layer for AI agents. OpenBrain combines semantic memory, vector search, tools, MCP, background workers, documents and human approval workflows into an internal AI operating system.
Node.js
PostgreSQL
pgvector
MCP
Fly.io
iKawn Mirror
AI-powered virtual try-on for physical retail. Built across the complete customer experience, including multimodal AI, kiosk UX, browser demos, lead qualification, retail workflows and deployment.
Multimodal AI
Retail UX
Computer vision
Deployment
iKawn Team
An agentic team layer for turning business context, tools and approvals into shared AI workflows. The focus is not a chatbot, but a system where agents can participate in how teams actually work.
Agents
Tool use
Approvals
Enterprise workflows
Forward deployed engineering means staying close to the workflow long enough for the software to become true.
Supporting Systems
The practical layer underneath
Not every useful system needs to be the headline. Some work matters because it makes existing businesses, tools and CMS infrastructure easier for agents to use safely.
Rameshwar BAS
A production business administration system covering payroll, attendance, invoicing, purchases, GST, TDS, banking workflows and document intelligence.
ProcessMCP
A simple ProcessWire plugin that makes existing ProcessWire sites easier for AI agents to access through MCP.
View on GitHubHow I Work
Understand, build, observe, iterate
I am comfortable moving between customer conversations, system architecture, database schemas, AI agents, API integrations, interfaces and production deployment.
01. Understand
Start with how the business actually works. The technology is secondary to the workflow that needs to become better.
02. Prototype
Move quickly enough to learn, but with enough architectural discipline that the prototype can become a production system.
03. Deploy
Get the system into the real environment, watch how people use it, tighten the edges and compound the useful parts.
Writing & Connect
Building AI systems that survive contact with work.
I write about Forward Deployed Engineering, agents, product and the business of turning emerging technology into something useful.