Week 3 · Day 21/30

Week 3 Capstone: Secure Agent Platform

Integrácia security, sandboxing, observability a design patterns

📅 2026-03-24 ⏱️ 8-10 hodín 📊 Security & Production
Celkový progres 70%

🎯 Cieľ dňa

Buildni secured, monitored agent platform s guardrails, approval flows, caching a full observability.

project core

📚 Study Resources

🛡️

OWASP — AI Security Guidelines

Comprehensive AI security resources od OWASP community.

reference
☁️

E2B Documentation

Cloud sandbox pre bezpečný AI code execution. SDK reference.

docs

💡 Key Concepts

Defense in Depth — Input filter → System prompt → Sandbox → Output filter → Monitoring → Alerting
Production Readiness — Security + Observability + Resilience + Cost control = production-ready AI system

🔧 Praktické cvičenie

Buildni end-to-end secure agent platform.

  1. Vezmi Day 14 multi-agent systém
  2. Pridaj input/output filtering (Day 15)
  3. Sandboxuj code execution (Day 16)
  4. Implementuj approval flow pre high-risk actions (Day 17)
  5. Pridaj OpenTelemetry tracing (Day 18)
  6. Implementuj semantic caching + circuit breaker (Day 20)
  7. Red-team celý systém a zdokumentuj findings
  8. Napíš security report

🧠 Self-Assessment Questions

  1. Aké sú top 3 LLM attack vectors podľa OWASP?
  2. Prečo Firecracker microVM je bezpečnejší ako Docker container?
  3. Kedy by si zvolil 90% confidence threshold vs 80%?
  4. Čo je semantic caching a kedy je efektívnejší ako exact match?
  5. Prečo je LoRA lepší ako full fine-tuning pre väčšinu use cases?
  6. Ako circuit breaker zabraňuje cascade failures?
AI System Engineer · 30-Day Study Plan · Generated 2026-03-03