Hebert NtseCloud & AI Security
Cloud & AI Security Engineer | AWS | Azure | GCP | Wiz | Terraform | Python

Security controls for cloud systems and AI applications.

I build cloud remediation workflows and AI runtime guardrails that turn risky inputs, findings, and model outputs into controlled, observable security decisions.

Open to Cloud Security, AI Security, DevSecOps, and Security Automation roles.
About

Security engineering across cloud infrastructure and the AI application layer.

I build practical security workflows across AWS, Azure, GCP, Wiz, Terraform, Python, and SQL. My work turns cloud findings and AI runtime risks into repeatable controls: inspect the signal, enforce policy, route exceptions, and leave a useful audit trail.

This portfolio shows how I think as an engineer: not just identifying cloud risk, but designing systems that help teams respond faster while keeping control over production change.

View certifications
Credentials

Certifications that back the cloud and AI security work.

CompTIA · CY0-001 CompTIA SecAI+

Securing AI systems, AI-assisted security operations, threat analysis, and responsible AI governance, risk, and compliance.

AWS AWS Certified Security - Specialty

Security architecture, identity, logging, monitoring, data protection, and incident response across AWS environments.

AWS AWS Certified Solutions Architect - Associate

Resilient cloud design, networking, IAM, storage, compute, and cost-aware architecture foundations.

HashiCorp HashiCorp Certified: Terraform Associate

Infrastructure as code, reusable modules, state management, secure delivery patterns, and repeatable cloud provisioning.

AI Security

Guardrails between users, models, tools, and sensitive data.

Featured Project · Google Cloud

Model Armor runtime security gateway

A reference implementation that inspects prompts before model inference and screens responses before release. Policy templates detect prompt injection, jailbreak attempts, sensitive data, malicious URLs, and unsafe content, while Cloud Logging preserves evidence for tuning and investigation.

Explore the architecture →
01User promptUntrusted input
02Model ArmorInspect and enforce
03LLMApproved inference
04Response scanPrevent data leakage
Hands-on Engineering

AI Projects & Labs

Lab 01 · Google Cloud · Multi-agent AI Workflow deployed to Cloud Run

Multi-Agent Zoo Tour Guide

A hands-on AI engineering lab that coordinates three specialized agents to welcome visitors, research zoo information, and format a tour response. Built with Google ADK and Python, using Gemini through Vertex AI.

Agent orchestration

Greeter, Researcher, and Response Formatter components use agent-to-agent handoffs and shared session state. A SequentialAgent coordinates research and formatting.

Identity & access

Authenticated Zoo MCP server integration using Google identity tokens, with Google IAM and service accounts configured for Vertex AI access.

Deployment & observability

Deployed on Cloud Run with Cloud Build and Artifact Registry. Used ADK tracing to inspect execution and tool calls, alongside Cloud Logging; troubleshot build permissions, Python dependencies, and deployment issues.

  • Google ADK
  • Python
  • Gemini / Vertex AI
  • Cloud Run
  • MCP
  • IAM & service accounts
  • Cloud Build
  • Artifact Registry
  • Cloud Logging
  • LangChain
  • Wikipedia

Deployment note: The agent workflow is deployed; some model requests encountered Vertex AI 429 quota errors.

View Architecture

Request flow

UserTour request
Greeter AgentWelcome & handoff
Session StateShared agent context
↓ SequentialAgent · research then format
Researcher AgentGather information
Zoo MCP Server / WikipediaResearch tools · LangChain integration
Response Formatter AgentCompose the tour response
UserFormatted response

The Researcher uses the Zoo MCP server / Wikipedia during its research step, then hands results to the Response Formatter through session state. MCP requests use Google identity tokens.

Google Cloud · supporting infrastructure
Gemini via Vertex AIModel inference for the agent workflow
IAM & service accountsVertex AI access permissions
Cloud Build → Artifact Registry → Cloud RunBuild, container storage & application runtime
ADK tracing & Cloud LoggingExecution, tool-call inspection & runtime logs
Cloud Security

Cloud automation built around controls, ownership, trust, and evidence.

Operating Model

Designed to automate clear decisions and escalate ambiguous risk.

01 Detect

Wiz controls identify cloud exposure patterns with structured context.

02 Validate

Python checks severity, ownership, resource type, account scope, and exception state.

03 Decide

Automation chooses remediate, notify, skip, or route for review based on guardrails.

04 Evidence

Every action records decision inputs, result, failure reason, and audit trail.

Contact

Let’s connect.

Get in touch about cloud security, AI security, and security automation opportunities.

Open to roles Cloud Security · AI Security · DevSecOps · Security Automation
hebertntse@gmail.com