RouteEye
BIAL · Production deploymentReal-time fleet infrastructure running in production at Bangalore International Airport.
High-frequency platform processing vehicle telemetry every ~5 seconds, streamed live to airport operations.
Hello, I'm Arjunagi Rehman
I build AI systems
that survive production.
Backend & AI Systems Engineer
Building production agents, real-time platforms and distributed systems — from 0→1 through production.
10+ Production systems
5+ AI deployments
BIAL Production deployment
4+ yrs Backend systems
AI agent, ready to answer questions
Your Question
Grounded in my projects & writing — answers include sources.
Selected Work
Systems that shipped
Production systems and experiments — what they do, who uses them, and how they're built.
Kalrav.AI
10+ live customersProduction AI agents deployed for real businesses.
No-code platform to create, deploy, and manage AI agents on any website — powering 10+ live businesses across WooCommerce, Shopify, and WordPress.
Agentic Dynamic UI
Open sourceInterfaces where AI returns interactive software instead of only text.
An agent that answers with live React components — a form when fields are missing, a chart when the trend matters — streamed over AG-UI with human-approved writes.
More projects & experiments
ChotU.AI
Multi-agent personal assistant built with LangGraph & OpenRouter. Supervisor agent routes to specialized sub-agents for task management, Google Calendar integration, daily planning, and personal data storage. Deployed on AWS ECS Fargate with React + Vite frontend on S3/CloudFront.
View case study
Wisp
Drink-spiking prevention platform combining portable spectroscopy hardware with a real-time mobile safety app. SOS alerts, live location sharing, and sub-3-second scan results.
View case study
AI/ML
Multimodal RAG ecai
Multimodal RAG platform on Azure. Documents uploaded to Blob Storage → text & images extracted → embedded via text-embedding-3-small → indexed in Azure AI Search → GPT-4.1 powered chat with contextual recall. Retrieval surfaces relevant text chunks + images together.
View case study
UCP Commerce Suite
End-to-end agentic commerce on the Universal Commerce Protocol: an AI shopping agent (Google ADK + Gemini), a FastAPI merchant reference implementation, and a Next.js storefront. The agent discovers the merchant from a single URL, links identity via OAuth, and completes real checkouts — all three services deployed and live.
GormAuditLog
Open-source Go library for ORM-level audit logging with GORM. Tracks create/update/delete events with minimal configuration overhead.
School Agents
Config-driven AI tutoring agent built with Google ADK and Gemini — YAML-defined agent, markdown prompts, dynamic tool loading, calculator and Google Search tools. Companion repo to the AI Agent System blog series.
BedRock Agent Runner
Local execution framework for AWS Bedrock agents — run and test Bedrock AI agents in a local development environment with multiple interface modes.
Gemini Multimodal Search
Image search powered by Google Gemini multimodal capabilities. Enables semantic search across images using AI embeddings and vector similarity.
Agentic-Dynamic-UI-CRM-example-
Agentic Dynamic UI — an AI agent that answers with live, interactive React components (forms, tables, charts, confirm cards) instead of plain text, streamed over AG-UI. CRM reference implementation.
ucp-agent
AI shopping agent (Google ADK + Gemini) that discovers, browses, links identity, and completes checkout on any UCP-enabled merchant — entirely through conversation.
ucp-merchant
Reference implementation of the Universal Commerce Protocol (UCP) — a FastAPI merchant layer that lets AI agents discover, authenticate, and transact with online stores.
ucp-demo-store
gormAuditlog
Go library that adds automatic audit logging to GORM apps — records INSERT/UPDATE/DELETE with user attribution, for MySQL and PostgreSQL.
gemini-multimodal-image-search
Multimodal image search — FastAPI app that indexes local images with Gemini embeddings into Qdrant and serves a browser UI for text- and image-based search.
Writing
How I think
Ideas I'm actively developing — on AI, interfaces, and where software is going.
Software Can Talk: How the Wall Between Humans and Machines Finally Broke A philosophical inversion of the oldest contract in computing — and the open-source agent stack behind /agent that lives the bet. Markdown-as-data, Haiku router + Sonnet responder, MCP from day one, ~500 lines. From Coders to Owners: AI Writes Code. Engineers Own Outcomes. Coding used to be the scarce skill. It is not anymore. An essay on how AI collapsed the engineering ladder, why ownership is the only thing left that matters, and what it means to be an engineer when the machine already writes the code. From CLI to AI: The Evolution of How Humans Talk to Software A deep narrative spanning punch cards, blinking terminals, GUI wars, the chatbot false start, and the LLM revolution — tracing how human-computer interaction evolved across eight paradigm shifts.
Toolbox
What I build with
Backend Go · Python · Java · TypeScript
Infrastructure AWS · Docker · Terraform · Redis · PostgreSQL
AI Google ADK · LangGraph · MCP · RAG · Bedrock
Real-time IoT · SSE · WebSockets · Kafka
About
I like building systems where software meets the messy real world. That has taken me from real-time infrastructure and IoT systems to production AI agents and agentic interfaces.
About me