Solution Architect with 10+ years of experience in enterprise IT architecture, project delivery, and technical leadership. Currently building AI-powered platforms and running them on self-hosted Kubernetes infrastructure.
AI/ML: LiteLLM, Ollama, RAG, Prompt Engineering, Vector Search (Qdrant), LLM Observability (Langfuse), Experiment Tracking (MLflow)
Languages: Python, Java, TypeScript
Cloud & Infra: Kubernetes (k3s), ArgoCD (GitOps), Traefik, cert-manager, Sealed Secrets, Prometheus, Grafana, Loki, Alertmanager
Architecture: C4 Model, Arc42, TOGAF, iSAQB
| Project | Description |
|---|---|
| SDLC Pilot | AI-powered SDLC automation — analyzes codebases, generates architecture docs (C4/Arc42), plans development work. 9-phase pipeline, 745+ tests, on-premises ready. |
Self-hosted sovereign AI platform running on k3s with full GitOps (ArgoCD App-of-Apps):
| Layer | Components |
|---|---|
| SSO | Authentik (OIDC/SAML) |
| LLM Runtime | Ollama + LiteLLM Proxy |
| AI Tools | Open WebUI, SDLC Pilot, Langfuse, MLflow |
| Data | Qdrant (Vector DB), PostgreSQL |
| Observability | Prometheus, Grafana, Alertmanager, Loki + Promtail |
| Infra | k3s, ArgoCD, Traefik, cert-manager, Sealed Secrets |
12 HTTPS endpoints, automated TLS, auto-sync with self-heal.
- Pipeline+LLM over Multi-Agent: Why single LLM calls beat agent loops (0% loops, 2x faster, 3x fewer tokens)
- Evidence-First RAG: Why LLMs should never guess about code — deterministic facts first, LLM synthesis second
- GitOps for AI Infra: Managing 10+ AI services with ArgoCD sync-waves and App-of-Apps
- On-Prem AI: Running LLMs, vector stores, and full observability on a single VPS
- Email: aymen@aymenmastouri.io