YK.
Back to projects

AI Operations Platform · 2026

IncidentOps AI

IncidentOps AI is a Cloudflare-native incident management and investigation application. It provides isolated workspaces for incident data, persistent incident timelines and operational memory, durable investigation workflows, authentication, and AI-assisted analysis through an integrated Copilot.

ROLE

Full-stack engineering · Cloudflare-native architecture · AI workflows

STACK

Cloudflare Workers · Workers AI · D1 · Durable Objects · SQLite · Cloudflare Workflows · JavaScript · HTML · CSS · Wrangler

OUTCOME

Built a production-deployed Cloudflare-native application combining authentication, workspace isolation, durable incident state, AI-assisted investigation, and persistent workflows. The architecture uses each Cloudflare primitive for a specific responsibility instead of relying on a traditional standalone backend.

SYSTEM / FLOW

01Request Layer

The browser sends application and API requests to the Cloudflare Worker, which acts as the main runtime and routing layer.

02Authentication & Workspace

Sessions, users, workspace membership, and ownership are validated through D1 before protected operations are allowed.

03Incident State

Each workspace keeps structured incident data, timeline events, investigation context, and operational memory in isolated Durable Object state.

04AI Copilot

Incident context is sent to Workers AI so the Copilot can analyze events, summarize incidents, and assist with investigation.

05Durable Investigation

Cloudflare Workflows coordinate longer-running investigation tasks that should continue reliably beyond a single request.

PROBLEM

What needed solving

Incident response information is often fragmented across messages, dashboards, notes, and individual sessions. AI assistance also becomes less useful when it cannot retain structured incident context or safely separate data between different teams and workspaces.

APPROACH

How I approached it

I designed the application around Cloudflare's platform primitives. D1 manages relational data such as users, sessions, workspaces, and memberships. SQLite-backed Durable Objects maintain isolated incident state and operational memory. Cloudflare Workflows handle durable investigation processes, while Workers AI powers incident analysis and Copilot capabilities. The application also supports anonymous demo workspaces and authenticated accounts for persistent cross-device access.

ENGINEERING

Implementation highlights

01Cloudflare-native serverless architecture
02Workspace and tenant isolation
03Authentication and session management
04Persistent incident state
05AI-assisted incident investigation
06Durable asynchronous workflows
07Operational memory and incident timelines
08Production deployment and environment configuration
09Automated testing

KEY DECISIONS

Trade-offs and choices

01

Cloudflare-native architecture

Used Workers as the application runtime so the API, authentication flow, AI integration, and platform services could operate within one deployment model.

02

D1 for relational state

Stored users, sessions, workspaces, memberships, and ownership relationships in a relational database where querying and constraints are important.

03

Durable Objects for incident state

Used SQLite-backed Durable Objects to provide isolated and persistent state for individual incident workspaces.

04

Durable investigation workflows

Used Cloudflare Workflows for operations that should survive beyond an individual request and continue reliably.

05

Workers AI integration

Kept AI analysis within the Cloudflare architecture and exposed it through the incident Copilot experience.

06

Demo and authenticated workspaces

Allowed users to explore the product immediately while authenticated accounts provide persistent workspace access across devices.

ENGINEERING SUMMARY

What this project is built on.

TECHNOLOGY

Cloudflare Workers · Workers AI · D1 · Durable Objects · SQLite · Cloudflare Workflows · JavaScript · HTML · CSS · Wrangler

ROLE

Full-stack engineering · Cloudflare-native architecture · AI workflows

OUTCOME

Built a production-deployed Cloudflare-native application combining authentication, workspace isolation, durable incident state, AI-assisted investigation, and persistent workflows. The architecture uses each Cloudflare primitive for a specific responsibility instead of relying on a traditional standalone backend.

NEXT PROJECT

Backend platform

Reading Platform

A backend-first reading platform with accounts, reading lists, reviews and administrative workflows.