I architect enterprise AI applications where Angular and TypeScript frontends meet RAG, MCP, tool execution, agent orchestration, human approvals, and governed AI runtimes.
- AI-native frontend architecture — streaming state, grounded citations, tool timelines, approval UX, context-aware interfaces, recovery flows.
- Enterprise Angular platforms — typed contracts, reusable SDK boundaries, scalable state, accessible operator workflows, backend-owned policy.
- Agent systems — provider-neutral runtimes, orchestration, MCP/tool integration, lifecycle governance, versioning, approvals, auditability.
- RAG and context architecture — inspectable retrieval, source provenance, context serialization, evaluation boundaries, safe backend integration.
- Human-in-the-loop systems — explicit approval gates, consequential-action review, failure recovery, revocation, and operator visibility.
- Platform engineering — NestJS/Node.js, Supabase/Postgres, Docker, Kubernetes, Playwright, CI/CD, security and observability.
| Project | Architectural proof | Primary signal |
|---|---|---|
| ngx-copilot-platform | Full-stack Angular copilot platform with a publishable SDK, backend-owned RAG/auth/approvals, SSE and API-key lifecycle | Flagship AI frontend platform |
| org-ai-force | Enterprise Angular + NestJS agent workspace with governance, RAG, MCP-style tools, browser workers and pilot operations | Enterprise AI architecture |
| frontend-ai-patterns | Reusable contracts and patterns for streaming, citations, tools, approvals, context, recovery and guardrails | AI frontend architecture patterns |
| agent-studio | Governed agent application factory with control plane, runtimes, workers, RBAC, lifecycle state and multi-surface publishing | Agent platform architecture |
| angular-ai-copilot-starter | Runnable Angular reference UX for Ask/Plan/Execute/Debug, RAG cards, tool visibility and approval states | Runnable frontend proof |
| ai-tools-cheatsheets | Open handbook for AI coding tools, MCP, engineering workflows, team adoption and governance | Open-source engineering leadership |
Enterprise Frontend Architecture
↓
AI-native UX
↓
RAG + Context
↓
Tool Calling + MCP
↓
Agent Orchestration
↓
Governed Enterprise AI Platforms
ai-tools-cheatsheets
Learn the AI engineering toolchain
↓
frontend-ai-patterns
Learn trustworthy AI interface patterns
↓
angular-ai-copilot-starter
Run the patterns in Angular
↓
ngx-copilot-platform
Integrate an application-grade platform
↓
agent-studio + org-ai-force
Govern and operate agent systems
- Agentic-Engineering-Playbook — six runnable systems covering provider gateways, RAG, orchestration, MCP, agentic Angular UI and browser automation.
- devdocs-forge-agent — local-first, provider-agnostic TypeScript documentation agent with reviewable output and source attribution.
- interviewOps — local-first Angular + TypeScript application demonstrating provider abstraction, schema validation, content architecture and ethics guardrails.
- shade-shifter — product engineering R&D spanning Flutter, BLE, ESP32, protocol design, CAD and hardware validation.
Frontend: Angular · TypeScript · RxJS · Signals · React/Next.js where appropriate
AI systems: RAG · MCP · agent orchestration · tool execution · approval workflows · OpenAI · Anthropic · Gemini
Backend & data: NestJS · Node.js · Supabase · Postgres · pgvector · Prisma/Drizzle · Redis
Delivery: Docker · Kubernetes · GitHub Actions · Playwright · observability · security-by-design
30 seconds: scan the six featured projects above.
3 minutes: open ngx-copilot-platform, frontend-ai-patterns, and agent-studio.
10 minutes: inspect architecture diagrams, trust boundaries, implementation-vs-demo notes, ADRs, tests and CI across the flagship repositories.
Frontend architecture is the interface. AI architecture is the system behind it. My work focuses on making the two operate as one product.