I'm Arun, a Full Stack AI Architect with 14+ years of building the systems that sit between a good idea and a production outage.
These days most of my work is about getting AI to behave in the real world: agents that call real enterprise APIs, models that have to answer inside a latency budget, and inference bills that someone in finance will eventually read. I'm currently a Solutions Architect & Full Stack AI Engineer at Creo Invent. Before that I designed multi-tenant SaaS platforms, enterprise integrations and cloud migrations at Infosys, PwC and TCS.
What I work on
Agentic AI. Multi-agent workflows, RAG, tool calling and MCP, LangGraph, structured outputs, evals and guardrails. Hands-on with Claude, OpenAI, Qwen, Llama and DeepSeek.
AI inference engineering. High-throughput, event-driven platforms for serving models and agents on AWS Bedrock and Azure, with a close eye on latency, observability and cost per request.
Forward-deployed engineering. Working inside client teams, learning the messy reality of their systems, and shipping AI that works there, not just in a demo.
Integration and platform. MuleSoft, Azure APIM and Functions, AWS Lambda and API Gateway, Terraform, Kubernetes, Kafka, and DevSecOps pipelines that don't need babysitting.
Full stack. Java/Spring WebFlux, Python, TypeScript/Node.js, React/Next.js, plus mobile with React Native, Flutter and SwiftUI.
Credentials
4x Claude certified by Anthropic (Architect – Professional, Architect – Foundations, Developer – Foundations, and the Claude Code Partner Badge), 9x Microsoft certified, and a MuleSoft Certified Integration Architect and Developer.
What I write about
Agentic AI, MCP, integration architecture, and the practical side of running AI in production. I also edit The Mule Blog.
Say hello on LinkedIn: linkedin.com/in/arundutta1
