Revamp dieu.dev as a platform/SRE/infra portfolio

[jonathandieu]

Tracking issue for rebuilding dieu.dev as a hireable portfolio. Living scope; refine as the brainstorm concludes.

Goal

A portfolio that makes a hiring manager believe I build and operate real production infrastructure. Target roles: platform engineering, SRE/reliability, cloud/DevOps infra, ML/AI infra. The strongest hook: the site runs on the very platform it documents, and proves it with live cost and reliability data.

Context / decisions so far

• The current portfolio repo is a fork of bchiang7/v4 with front-end project content. Redo from scratch; this is a repositioning from "front-end dev" to "platform engineer," not a reskin.
• Not a designer: use a clean, minimal off-the-shelf theme. Substance (real architecture, incident writeups, live system data) carries it, not visual flourish.
• Build and evolve in public. The roadmap and changelog become site content (this also satisfies the "centralized roadmap" ask).

Content architecture

The site is evidence, and most of it already exists from real work.

Receipts (core credibility):

• Platform overview: one architecture diagram + short narrative (clusters, GitOps via ArgoCD/Kargo, Cilium gateway, workloads).
• Deep-dive case studies (problem -> decision -> implementation -> outcome, linking real ADRs/PRs):
  • Debugging a 4.5-day load-balancer reconcile loop (the 422 saga)
  • Migrating Envoy Gateway -> Cilium Gateway API
  • Multi-cluster failover with Cloudflare Workers
  • GitOps from zero: terraform + ArgoCD + Kargo + dieuctl

Live exhibits (added over time):

• Live cost page (Public cost-transparency page: run cost + per-session demo cost #32)
• Load-test button + rate limiting (Cluster-layer rate limiting for the load-test button (scoped Envoy GatewayClass) #26)
• On-demand GPU / serving demo
• Status / reliability page (failover)

Building-in-public spine:

• Public roadmap + changelog auto-generated from GitHub issues/PRs

Basics: concise about/contact/CV, links to the repos as source-of-truth receipts.

Open decisions

• Site concept direction. Options under review in brainstorm: A control-plane (homepage as live ops dashboard), B engineering-docs (clean, content-first), C interactive architecture (clickable live system diagram). Current lean: B as the readable base with A's live tiles on the homepage, C as a later showcase page.
• Hosting/tech: self-hosted on the DOKS cluster via GitOps (most on-brand) vs external static host with live cluster data. Static site generator (Astro/Hugo) to minimize front-end burden. Where the code lives (dieu.dev repo vs reuse portfolio repo).

v0 slice (build-in-public starting point)

Landing + platform overview + one case study (start with the 422 incident) + the auto-generated roadmap, deployed on the cluster. Everything else lands as visible PRs.

Related

• dieubernetes: Public cost-transparency page: run cost + per-session demo cost #32 (cost page), Cluster-layer rate limiting for the load-test button (scoped Envoy GatewayClass) #26 (load-test/rate-limit), whereis.dieu.dev: decide requester auth/identity approach #16 (whereis.dieu.dev), Configure dieu.dev domain with Cloudflare #5 (domain)
• terraform: feat: workload ApplicationSet + charts #14 (Cilium migration), Cross-cluster networking for unified Grafana + cost dashboard #15 (LB + failover strategy)
• Not yet filed: on-demand GPU nodes, handoff-docs standardization (to be linked from the umbrella roadmap issue)

[jonathandieu]

Homepage direction (leaning): content-forward A+B blend. Positioning headline leads, a compact live status strip under it (clusters up, monthly cost, uptime), then selected work. Full dashboards live on the dedicated cost/status pages. Next: overall page/nav structure.