Week 4 · Day 24/30

Production Deployment

Scaling AI, load balancing, A/B testing models, canary rollouts

📅 2026-03-27 ⏱️ 6-7 hodín 📊 Mastery & Portfolio
Celkový progres 80%

🎯 Cieľ dňa

Naučiť sa deployovať AI systémy do produkcie: scaling, A/B testing modelov, canary rollouts, a monitoring.

core practice

📚 Study Resources

💻

Temporal — Durable AI Agent Tutorial

Hands-on: buildni durable AI agenta s Python a Temporal. Prežije crashes.

tutorial
☸️

KServe — Model Serving on Kubernetes

Serverless model serving: autoscaling, canary rollout, A/B testing, monitoring.

docs
🏗️

Temporal — Multi-Agentic AI Architecture

State prežije crashes, automatic retry. Production multi-agent patterns.

article

💡 Key Concepts

Canary Deployment — Nový model pre 1-5% traffic. Monitor metriky. Postupne zvyšuj alebo rollback.
A/B Testing Models — 50/50 split. Štatisticky dokaž ktorý model je lepší pre business outcome.
Autoscaling — Horizontálne scaling podľa request volume. Kubernetes HPA + KServe.
Durable Execution — Temporal: agent state prežije process crashes. Idempotent activities, automatic retry.

🔧 Praktické cvičenie

Deployni AI agent s canary rollout.

  1. Containerizuj agenta z predchádzajúcich dní
  2. Nastav Docker Compose alebo K8s pre deployment
  3. Implementuj health checks a readiness probes
  4. Pridaj canary deployment config (2 verzie modelu)
  5. Monitor metriky a simuluj rollback scenario