Hi, I'm Murali.
RSS FeedI'm a Senior AI Platform Engineer, focused on building agentic AI systems that hold up in production. My work spans multi-agent orchestration, secure runtimes, memory and context management, RAG, MCP-based tools, observability, evaluation pipelines, and model routing across cloud platforms (AWS Bedrock, GCP Vertex AI etc), with EU data residency designed in. I draw on more than a decade of experience in cloud, security, and infrastructure to take AI systems from early prototypes to secure, resilient, and cost-aware platforms. I'm continuing to deepen my expertise across AI engineering and AI platform engineering, and I write here about what I'm building and learning along the way.
Featured
-
Rearchitecting an Agentic App: From ClickOps Across Five Clouds to an AI Platform
An agentic app grew out of an MVP with components scattered across five clouds, all provisioned by hand. Here's how I rearchitected it — Terraform first, agent-level observability second, consolidation last — and what it taught me about building a harness for AI agents.
-
I Let an AI Agent Run My FIFA World Cup 2026 Fantasy Squad
The World Cup starts next week. 736 players. 48 teams. 8 matchdays. So I built an AI agent that uses Poisson models, integer linear programming, and a bounded Claude layer to pick the mathematically optimal fantasy squad.
-
From API Tokens to AWS Bedrock: My Real-World Journey Configuring an AI Agent
I set up a personal AI agent and spent weeks fighting cost overruns, local model disappointments, and AWS throttling. Here's the real story — every wrong turn, every fix, and the multi-model architecture that finally worked.
-
How I Use AI Agents to Analyse Costs Across Large AWS Environments
Managing AWS costs across dozens of accounts is painful. Here's how I use AI agent skills to trace billing spikes to root cause, find waste, and recover thousands per month — without clicking through the console.
Recent Posts
-
What It Takes to Run Agentic Systems in Production: A Systems Guide
A map of the runtime, memory, tools, knowledge, security, observability, evaluation, and infrastructure required to turn an agent loop into a production system.
-
Building Multi-Architecture Container Images Using Kaniko and GitLab CI
How to build multi-architecture container images using Kaniko and GitLab CI, targeting both amd64 and arm64 for AWS Graviton on ECS Fargate — with native runners per architecture and manifest-tool to combine them.
-
Building a DevSecOps Pipeline Using Open Source Tools
A complete DevSecOps pipeline using open source tools — Dependency-Track, CodeQL, Gitleaks, Checkov, Trivy, and DefectDojo — integrated into GitLab CI/CD for a Terraform + Node.js + GCP Cloud Run stack.
-
Automating Certificate Management using GKE's Managed Certificates
How to use GKE Managed Certificates to automate TLS certificate lifecycle for applications running on Google Kubernetes Engine.
-
Automating Certificate Management for Kubernetes using Cert-Manager and Let's Encrypt
How to automate TLS certificate lifecycle management for Kubernetes applications using cert-manager and Let's Encrypt free certificates.
-
SSO for Managed Kubernetes Services (Part 1)
Setting up unified OIDC-based authentication across EKS and GKE using Okta as a single source of truth, eliminating duplicate identity management across cloud providers.