PaulZanbaka
If you can imagine it,
I can build it.
AI product engineer, enterprise automation architect, and cloud/cybersecurity engineer. I turn hard ideas into working products across agentic AI, pentesting, audit readiness, trading systems, and cloud SaaS.
About
Passionate, self-driven engineer who thrives on difficult problems — turning ambitious ideas into working systems across AI, cybersecurity, cloud infrastructure, and enterprise automation.
Hands-on across software development, AI product integration, penetration testing, vulnerability management, database administration, and hybrid cloud estates — with the leadership experience to take products from concept through production.
Known for combining deep technical range with practical execution: agentic multi-agent workflows, secure API design, CVSS-prioritized findings, human-in-the-loop safety, and comprehensive testing across sandbox through QA before production deployment.
VCP
PMP
Security+
AI / LLM engineering
Agentic architecture, MCP tooling, Claude Code workflows, RAG, structured output, and eval gates — applied in shipped products, not slides.
Multi-agent system design across hub-and-spoke and orchestrator–subagent topologies; task decomposition and planning, dynamic replanning and ambiguity handling, in-app self-healing agents, and human-in-the-loop approval gates.
Shipped in zaiaudi — task-decomposed subagents for ingestion, retrieval, and evidence validation.
MCP client/server development, tool-boundary design that prevents reasoning overload, custom tools, Skills, and subagents; infrastructure monitoring, automated recovery and alerting, and ServiceNow incident / request / catalog integration.
Shipped in 1Ops — ServiceNow work turned into reusable agent workflows at Rollins.
CLAUDE.md hierarchy, custom slash commands and Skills, CI/CD and headless invocation, and local AI implementation — the developer platform tuned so agents build the way the team builds.
Applied across zaiaudi and PentestMyApps — CI/CD and headless invocation drive the build-and-verify loop.
JSON-schema enforcement, few-shot technique, validation and retry loops, and structured-output handling for both feature development and data analysis.
Shipped in zaiaudi control mapping and readiness scoring, and PrintMoney AI signal recommendations.
RAG architecture — chunking, embedding, hybrid retrieval — plus prompt caching, Batch API workload design, long-conversation compaction, and session memory.
Shipped in zaiaudi — a RAG pipeline for document ingestion and search across the evidence set.
Task-specific model routing, LLM fundamentals, cost and token management, model hardening and health monitoring, and skills development and customization.
Shipped in 1Ops — task-specific routing across incident, fulfillment, and health workflows.
Human-in-the-loop safety, secure API/token and identity, secrets, and key management, guardrails and safe deployment, code review, and multi-agent build verification.
Shipped in PrintMoney — hard risk guardrails, paper mode, kill-switch controls, human-in-the-loop approval.
Production Claude application design, configuration management, AI adoption roadmap development and execution, team collaboration and training, executive reporting and dashboard visibility.
Shipped across zaiaudi executive dashboards, TrinityX risk posture, and 1Ops executive insight.
Comprehensive manual and automated testing across Sandbox, Development, Test, and QA; verification and refinement pipelines; audit, vulnerability management, and penetration testing before anything reaches production.
300+ automated tests in zaiaudi; audit, vulnerability management, and penetration testing before every production deployment.
Also in the toolkit
Core skills
Software, automation, cyber tooling, cloud SaaS, and leadership — applied end to end from idea to production.
Python, PowerShell, ASP, C#/.NET, JavaScript, Bash, SQL, REST APIs, automated testing, OWASP, system design, and software architecture.
Azure Logic Apps & Automation, SCORCH, PowerShell, ServiceNow orchestration, runbook engineering, self-healing recovery, request fulfillment, and operational dashboards across hybrid estates.
Penetration testing platforms, systems hardening, vulnerability management, CVSS prioritization, incident response, and risk management — built as shipping applications and tooling, not paperwork.
AWS — VPC, IAM, EC2, S3, Lambda, RDS, Athena, CloudWatch, CloudTrail, CloudFormation, Route 53, Security Groups. Azure — Entra ID, Azure Arc, and hybrid Windows/Linux estates.
0-to-1 delivery, multi-quarter initiatives, cross-functional influence, hands-on problem solving, mentoring and up-leveling teams, tradeoff decisions, and operational reliability.
Linux, Windows, SQL DBA, VMware (vSphere, NSX), MECM, SCOM, LogicMonitor, Active Directory, Group Policy, DNS, DHCP, SMTP, Apache, and IIS.
First instinct is to dig in — isolate the failure, map the system, and ship a durable fix. Comfortable in the hard rooms: outages, ambiguous requirements, and cross-team blockers.
Georgia Tech Applied Physics trained me to model reality before coding it — measure, simplify, then build systems that hold under load. Precision and curiosity still drive every architecture decision.
I take a rough concept and drive it to a live subscription or enterprise system — architecture, agents, security, and operations included. If you can imagine it, I can build it.
Enterprise service catalogs & APIs
Multi-tenant SaaS · SSO · RBAC
VPC · IAM · EC2 · S3 · Lambda · CloudWatch · CloudTrail · Route 53
AI pipelines · FastAPI · automation
Azure Logic Apps & Automation · SCORCH · PowerShell
Offensive security tooling
AI-engineered products
Three production products, each with its own lane. Every screenshot below is live product proof — click through and use them.
"Audit-ready, automatically. Walk in already knowing the result."
Built 0-to-1 on an agentic architecture — task-decomposed subagents for ingestion, retrieval, and evidence validation — across 50+ Python modules and 300+ automated tests. A RAG pipeline ingests policies, runbooks, tickets, and evidence logs; structured-output validation maps every artifact to ISO 27001:2022 controls, scores the gaps, and hands you a sample for every missing control.
"Your apps have holes. We find them before someone else does."
Orchestrates 18 scanners and an 8,000+ template engine as tool integrations, producing structured, CVSS-prioritized findings mapped to OWASP, NIST, PCI DSS, SOC 2, and ISO 27001. Coverage spans SSL/TLS, headers, ports, SQLi/XSS, APIs, GraphQL, JWT/OAuth, subdomain takeover, secrets, email security, and Nuclei templates.
"Watch the floor. Trade the floor."
Integrates with brokerages, user-configured tool APIs, financial news, and live market feeds. The central Decision Desk is an AI-integrated dashboard with chart indicators and gauges, back-tested strategies, and structured-output AI signal recommendations — with TraderTV docked right inside it.
Unified vulnerability management and automated remediation. Agentic orchestration correlates findings across Tenable, cloud-native security services — Intune, AWS Inspector, Defender — and open-source scanners, then drives remediation workflows, remote server management, and executive risk-posture dashboards.
AI-powered infrastructure management, visibility, and automation. Converts repetitive ServiceNow incidents, request fulfillment, operational tasks, identity and licensing cost-savings, security audit support, and infrastructure health data into reusable agent workflows with task-specific model routing — and executive-visible insight.
Experience
From accredited DoD systems to multi-region SaaS and an independent AI product studio.
Designs, ships, and operates live products end to end across audit readiness, pentest automation, security operations, and risk-governed trading workflows.
Conceived and built 1Ops. Automated recovery, monitoring, ServiceNow work, identity signals, and executive reporting across large on-prem, Azure, AWS, and endpoint estates. Led the recovery of two datacenters, 2,000+ servers, and VIP desktops during the CrowdStrike incident.
Built the ISMS from zero. Led ISO 27001 certification (2023) and surveillance (2024); advising on 2026 recertification. Architected secure multi-region AWS SaaS for research institutions including Emory, Harvard, Rockefeller, Yale, Oxford, Max Planck, Cambridge, and Institut Pasteur — per-tenant isolation, WAF-fronted ingress, and EU data-handling requirements.
300+ orchestration runbooks across a 2,200-server Windows/Linux/AIX estate — application recovery, one-click provisioning, Active Directory audit, and ServiceNow request fulfillment. Built a C#/PowerShell service catalog integrated with HPSM for order tracking and reporting. Self-remediation and intelligent routing cut after-hours on-call alerts ~88%.
Led software engineering team automating IT service desk incidents and tasks. Drove RMF inspection readiness, vulnerability-management decision support, and command dashboards adopted by U.S. Army IT Command.
Programmer Analyst → Software Engineering Manager → Senior Operations & Engineering Manager. 20+ accredited DoD applications across Airborne School, Ranger Training Brigade, and Officer Candidate School. Best-in-class DIACAP (2014) and CCRI (2015) scores. $2M+ annual savings via data-center consolidation across five sites. Spot Award from the Commanding General, U.S. Army 7th Signal Command.
Foundation
Two degrees and a consolidated certification foundation across security, data, virtualization, service management, web engineering, and project delivery.
Georgia Institute of Technology — High Honors
Auburn University — GPA 4.0, Highest Honors
Certifications
Let's build
If you can imagine it, I can build it — AI products, cyber tooling, and cloud SaaS that ship.
