The AI SRE race will not be won by the agent that diagnoses fastest. It will be won by the system that operators trust enough to grant write access.
A personal note on where AI for operations is actually heading.
Over the past year a new category filled up fast. Depending on how you count, there are now more than a dozen credible tools that call themselves AI SREs. I have watched the space closely, partly because we are building in it, and partly because the speed of convergence is genuinely interesting.
Here is what nearly all of them do. They connect to your telemetry, your code, and your incident tooling. They correlate logs, metrics, and traces. When an alert fires, they form hypotheses, test them against the evidence, and post a likely root cause into Slack, often in under a minute. This is real progress. A few years ago none of it worked. Today most of it does.
Phase one is real
Diagnosis is real progress. But it is just phase one.
When something breaks in production at an odd hour, the person on call has to do three things at once: understand what is happening, decide what to do about it, and be able to explain all of it the next day. Most AI incident tools help with at most one of these. They either give you more data to read, or they take action you cannot see and cannot account for afterward.
We spent the last several months building Nova's investigation engine around that gap. This post is about how we designed it, the models we borrowed from, and the trade-offs we made along the way.
Six months in as platform lead and you have a spreadsheet you haven't shown your manager. Eleven Kubernetes clusters. EKS for production. GKE for the ML team. Two on-prem clusters behind the firewall that predate your tenure. A handful of Kind clusters developers spun up locally. Each one has its own deployment pipeline, its own credentials rotation process, its own way of answering "is this service healthy?"
Your team isn't building features anymore. You're maintaining eleven slightly different versions of the same tooling.
An AI platform engineer alongside a programmable production foundation.
AI agent demos tend to look easy. Ask a question, call a tool, show an answer.
Production infrastructure is less forgiving. Context changes, tools fail, permissions differ, and an apparently reasonable action can have real consequences.
We have spent over a year building Nova, our AI platform engineer, around those realities. This is what we believe matters when an agent moves from answering questions to doing engineering work.
Platform teams carry operational knowledge that does not transfer easily. The debugging instincts, the service interdependencies, the deployment quirks: they accumulate over years and live in a few people's heads. When those people are unavailable, the gap shows.
We built Nova to put that knowledge into a system you can query. This post covers what the architecture looks like and what we learned building AI that actually operates infrastructure.
Enterprises are discovering they can run powerful AI models on their own infrastructure, but building production AI infrastructure is significantly harder than application deployment.
This post breaks down the six interconnected systems required, why specialized small models outperform foundation models for enterprise use cases, how emerging AI agents are changing the economics, and the engineering trade-offs at every layer.
📖 About This Guide
This is a comprehensive technical deep-dive. We explore the complete AI infrastructure landscape, from why enterprises build their own platforms to the six pillars required and the open-source technologies available.
You want the simplicity of "push code, get a live URL" (the developer experience Vercel pioneered) but with full control over your deployment, infrastructure, and compliance. This guide shows you how to build that experience on your own AWS infrastructure using AstroPulse and open-source tools: kpack, cert-manager, external-dns, and nginx-ingress.
You'll build a production-grade platform that delivers Git-push deployments with automatic TLS certificates, preview URLs, and complete observability, all running on infrastructure you own and control. Unlike hosted PaaS platforms, you'll be building on Kubernetes with full deployment control. That means you can run any workload: microservices (with or without public endpoints), stateful databases, WebSockets, long-running background jobs, AI/ML model training and serving, or traditional web applications in any language. You get the simple developer experience with complete architectural control.
How operations work: The infrastructure industry is moving toward an agentic era: AI agents autonomously handling complex workflows (MCP, A2A, multi-agent orchestration). We're heading toward infrastructure that self-configures, self-heals, and self-optimizes. We're not there yet, but Nova brings you AI-assisted operations today with human-in-the-loop. Day 1 (this guide): You build the platform. Day 2 (ongoing): Nova analyzes issues, diagnoses problems, recommends fixes, and you approve. As AI matures, more becomes autonomous.
📖 About This Guide
This is a comprehensive, production-ready blueprint. We cover everything from architecture to production deployment with complete working examples, security, compliance, and troubleshooting.
⚡ Want the fast track? Jump to the setup walkthrough (platform deploys in 30 minutes)
Platform engineering represents the natural evolution of DevOps and SRE principles, but it faces a fundamental challenge: how do you scale platform expertise across an entire organization without requiring every developer to become a cloud expert?
This is where Nova comes in: your AI platform engineer that makes infrastructure accessible to everyone through natural conversation.
The Platform Engineering Evolution
DevOps broke down silos → SRE brought engineering rigor to operations → Platform Engineering created self-service infrastructure → Nova makes platform engineering conversational and accessible to everyone.
The Platform Engineering Challenge: Scale vs. Expertise
Platform engineering promised to solve the "you build it, you run it" scaling problem by creating Internal Developer Platforms (IDPs). But even the best platforms face fundamental limitations:
👥
Expert Bottlenecks
Platform teams become the new constraint, since everyone depends on their expertise
📚
Documentation Decay
Complex systems require constant documentation that quickly becomes outdated
🧠
Context Loss
Critical operational knowledge lives in tribal knowledge, not systems
⚙️
Cognitive Load
Developers still need to understand infrastructure concepts to use platforms effectively
The Core Issue
We've built self-service platforms, but we haven't solved the underlying problem of democratizing platform engineering expertise.
Nova is an AI platform engineer that helps you manage infrastructure through natural conversation. Ask questions, get answers, generate configurations, and troubleshoot issues, all through simple chat.
What makes Nova different:
Works with your existing tools (Slack, GitHub, AWS, Terraform, Kubernetes, and more)
Available however you want to work: browser, self-hosted, or in your editor
Nova's power comes from its extensibility. Connect the tools you already use:
Available Skills:
Cloud Providers: AWS, Google Cloud, Azure cost calculations and resource management
Communication: Slack integration for team collaboration
Development: GitHub for code search, issues, and PRs
Infrastructure: Terraform and Helm configuration generation
Kubernetes: Cluster management and troubleshooting
And more: Add any MCP server for custom integrations
Bring Your Own Tools
Nova Direct includes the built-in MCP Marketplace for custom integrations. Nova Connect works through standard MCP-compatible clients, so teams can bring Nova into existing editor and CLI workflows without running the server locally.
The evolution from DevOps → SRE → Platform Engineering → AI-Assisted Platform Engineering represents more than technological progress. It is about democratizing expertise that has historically been scarce and expensive.
Traditional Model
Small teams of platform experts serve entire organizations