Build an internal EC2 fleet dashboard that gives an engineer a fast, at-a-glance view of every running instance across all their AWS regions, so they can spot what's running, where, and for how long without digging through the AWS console. Trigger it as a web page served on an authenticated route (keep it private to the space or team). On load, the page should fetch live instance data from a backend endpoint in the same workflow, and offer a manual refresh. Build these pieces at a high level: A data step that scans across regions and returns each instance with its state, type, platform, IPs, region/AZ, launch time, attached volumes, security groups, and tags. Use the AWS EC2 API (or your cloud provider's equivalent) via a connector so no credentials live in code. A dashboard UI that renders the instances in a sortable, searchable, filterable table (filter by state and region; default the sort to most recently launched) with summary counts and expandable rows for full detail. Aim for a clean, calm, production-grade look. An on-demand endpoint that fetches a chosen instance's console/system log. An AI endpoint that takes a console log and returns a one-sentence plain-English summary of what's happening, surfaced behind an "Explain with AI" button in the log panel. Use an LLM API such as Anthropic Claude — or your preferred model provider. Keep all endpoints authenticated, credentials in connectors, and the whole thing read-only against your cloud account. Tools used: AWS EC2 API (via AWS role connector), Anthropic Claude Messages API