How this website is built
The build pipeline, shared platform, and delivery patterns this site runs on.
Static generation, no runtime
A small Crystal generator in src/sitegen.cr reads YAML content and renders ECR templates into a fully static dist directory. The deployed site has no runtime, no database, and no per-request server code—every URL is a pre-built file, which keeps the operational surface area near zero.
Versioned design system on a CDN
Tokens, shell, and primitives are pinned under static.michaelj43.dev/v1/ (published from the static-assets repo): Terraform-defined S3 plus CloudFront with Origin Access Control, immutable caching for versioned assets, and a short-lived cache policy on index.html. One centralized visual system, many consumer subdomains, no re-bundling.
Decoupled auth shell
The top navigation is injected at runtime by m43-auth-header.js, which resolves session state against the auth platform via api.michaelj43.dev. The portfolio stays fully static—sign-in concerns never bleed into the build—so the HTML remains straightforwardly cacheable.
Telemetry to a shared API platform
Page events post to a versioned /analytics/events endpoint on api.michaelj43.dev, backed by shared-api-platform: API Gateway, Lambda, and DynamoDB, deployed via GitHub Actions with AWS OIDC. CI gates the OpenAPI contract (Spectral lint and drift checks), and CloudWatch alarms watch production. This site identifies itself as portfolio-pages via data-m43-app.
Terraform-defined, OIDC-delivered
The runtime surfaces this page depends on—the static-assets CDN and shared-api-platform API—are Terraform-defined with S3 plus DynamoDB-locked remote state. Their deploys, plus iac-builder, card-game, kubernetes-mono-app, and local-ai-worker, authenticate to AWS via GitHub OIDC into scoped roles, so long-lived cloud credentials do not live in repository secrets.
CI/CD with ephemeral PR previews
GitHub Actions builds and publishes on every push to main with single-flight concurrency. Each pull request gets its own preview folder on gh-pages (for example preview/pr-42/), surfaced via a sticky bot comment and removed when the PR closes—the same ephemeral-environment idea as iac-builder (PR-isolated Terraform on AWS) and card-game (preview multiplayer stacks).
Skills
YAML-driven technical depth and workplace collaboration skills, grouped for quick scanning.
Technical
Languages
- Python
- TypeScript
- Java
- Go
- Groovy
- Kotlin
- PowerShell
- Bash
- Rust
- HCL (Terraform)
- Crystal
- YAML
Cloud and infrastructure
- AWS
- Microsoft Azure
- Infrastructure as Code
- Multi-region deployment
- Virtual machines and containers
- CDN and static asset delivery
DevOps and automation
- CI/CD pipeline design
- Release management
- Automated test workflows
- Developer enablement
- Process automation
- Cross-team delivery
Product and platform engineering
- TypeScript applications
- Cross-platform desktop tools
- API integration
- Auth flows
- Shared UI systems
- Static site generation
Systems and observability
- Rust async tooling
- Go concurrency
- Streaming pipelines
- New Relic dashboards and alerts
- Slack alert integration
- Performance-minded services
Security
- Identity and access management
- Secrets management and rotation
- Encryption at rest and in transit
- CI/CD & supply chain security
- IaC policy & guardrails
- Vulnerability management
Code & dependency scanning
- SonarQube / SonarCloud
- Mend (WhiteSource) & SCA
- SAST & quality gates in CI
- Repo secret scanning
- Image & binary scanning
- SBOM & dependency policy checks
Workplace & collaboration
Productivity suites
- Microsoft 365
- Google Workspace
- Spreadsheet analysis & reporting
- Presentation design & narrative
Delivery & agility
- Agile, Scrum, or Kanban
- Iteration planning
- Backlog refinement
- Release coordination
Collaboration & leadership
- Stakeholder alignment
- Technical writing & runbooks
- Mentorship & knowledge sharing
- Data-informed prioritization
Work tracking & communication
- Jira & Azure DevOps
- Confluence & internal docs
- Slack & Microsoft Teams
- Incidents & change requests
AI productivity tooling
- ChatGPT, Copilot, & Gemini for drafting & review
- Cursor, Windsurf, & Claude for agentic work
- Structured prompting & evaluation
- Guardrails & safety framing
- AI-assisted workflows
Distributed team operations
- Async-first communication
- Offshore & cross-timezone handoffs
- Meeting hygiene & agendas
- Decision logs & follow-ups
Selected work
-
Echo
- Tauri desktop apps
- API client UX
- Auth workflows
- CI/CD and PR previews
Cross-platform desktop app for exploring HTTP APIs—collections, environments, auth, and request/response views—built with Tauri. The public site at echo.michaelj43.dev is the marketing and documentation surface, with automated deploys and PR previews for contributor-friendly iteration.
-
card-game
- React + TypeScript UI
- Multiplayer REST + WebSocket APIs (AWS)
- WebRTC data channel + signaling
- DynamoDB + Lambda room services
- Terraform + custom domain hosting
Browser card table with many YAML-driven games, AI seats, and online play. An AWS multiplayer stack handles rooms, JWT-gated signaling over WebSocket, and optional TURN scheduling; peers sync authoritative table state over WebRTC while Terraform defines the infrastructure. The live app is on a custom domain so friends and reviewers can try builds without cloning.
-
kubernetes-mono-app
- EKS-oriented Kubernetes
- Argo CD GitOps
- Terraform (VPC, EKS, OIDC)
- Kustomize + ALB Ingress
- GitHub Actions (CI, images, bootstrap)
- CloudNativePG and Redis patterns
Portfolio mono-repo with a Go API, static portal (apex landing plus public /status for Argo health), GitOps Application CRs, Terraform for VPC/EKS and GitHub OIDC roles, Kustomize overlays with ACM discovery for TLS, and Actions for tests, GHCR images, and Argo bootstrap—oriented toward reproducible bring-up on AWS.
-
iac-builder
- React + Vite wizard UI
- Optional OpenAI BYOK assist (auditable prompts)
- Optional encrypted AWS profiles + VPC/subnet discovery
- Terraform-style generation + code preview drawer
- Go API with SQLite and preview pipeline
- Docker Compose; AWS serverless hosting (GitHub Actions)
Guided Infrastructure-as-Code generator: a React wizard fronts a Go API that produces previewable Terraform-style output in a code drawer, with sensible defaults, optional encrypted AWS credential profiles for discovery, and optional OpenAI BYOK suggestions under operator-controlled guardrails. Local runs use Docker Compose; production paths cover S3, CloudFront, and Lambda with documented Terraform and CI deploy/destroy workflows.
-
spiffy-trader (Github)
- External API integration
- Model-assisted evaluation
- Automation
- Reproducible workflows
Kalshi trading experiments that combine model-assisted signals with structured evaluation and clear boundaries between research and production concerns. The repository tracks orchestration, API usage, and how results are recorded so ideas stay reproducible and reviewable over time.
-
local-ai-worker
- Rust async
- Local AI runtimes
- Service boundaries
- Performance-minded tooling
Rust worker for local or self-hosted AI workflows, oriented toward performance and clear boundaries when talking to model runtimes. It shows how I lean on the type system and async tooling in Rust to keep integration code small and testable.
-
adaptive-pipe (Github)
- Go concurrency
- Streaming pipelines
- CLI and service building blocks
- Small interfaces
Go project for adaptive streaming or piping between components, useful for composable CLI and service edges. The code favours small interfaces, straightforward concurrency, and building blocks that are easy to reason about in larger systems.
