AI Operations Engineer
Moneta Group
- Operate and improve AI workflows inside a wealth management firm.
- Focus on systems people actually use, not pilots that die in a slide deck.
$ whoami
Zachary Montague
AI Operations Engineer
I put AI into the work you already do
Zac Montague builds the thing and leaves it running, or sits with your people until they can run it themselves. Most clients want some of both.
- Build or teach. Usually both.
- Same ops habits from SRE and regulated cloud: it has to survive Tuesday.
- Client work goes through Vadrel. This site is me.
Working with companies through Vadrel
Terminal below. Try help, neofetch, or cd experience. Press ` to focus.
identity.img verified
About Zac
Zac Montague is an AI operations engineer in the St. Louis area. He puts AI into work companies already do: records, follow-up, the stuff that lives in email and spreadsheets. Client work goes through Vadrel. Before that he was an SRE at Avast and a cloud engineer at Saint Louis University, including the medical school.
He spent years in production operations, where automated systems fail under load and someone has to decide when a human takes over. That is the same problem as putting AI into a real company.
Today he builds and teaches those systems. Not strategy decks. Something in production.
Experience
Production operations, governed automation, and security review since 2019, with a current focus on AI workflow governance.
Site Reliability Engineer
Avast (Gen Digital)
- Operated production Kubernetes platforms with strict deployment controls, cutting rollout errors by 40%.
- Built Datadog observability with custom SLIs, SLOs, and anomaly detection.
- Automated incident triage in Python with PagerDuty, routing by severity and escalating to humans at defined boundaries.
Cloud Engineer
Saint Louis University School of Medicine
- Designed HIPAA-governed cloud environments with encryption, access controls, and audit trails for medical and research data.
- Standardized Terraform and Ansible so production changes were repeatable and reviewable.
- Provisioned HPC cloud capacity for data-intensive research workloads.
Cloud Engineer
Saint Louis University
- Architected multi-cloud AWS and Azure infrastructure with controlled change management.
- Built CI/CD pipelines and auto scaling so services stayed stable under demand spikes.
Founder
Nomad Cybersecurity Group Ltd
- Founded an AIFC-registered company in Astana. Concluded active management in June 2026. That work is closed.
Expertise
AI in the work you already do
Records, follow-up, intake, the messy stuff in email and spreadsheets. Built into the workflow, not parked in a chatbot nobody opens.
Build or teach
I leave you a running system, or I sit with your people until they can run it. Most engagements are some of both.
Production operations
Kubernetes, observability, incident automation, and infrastructure that survives audits and outages. Same habits, now applied to AI.
Human review when it matters
Gates before anything customer-facing. Escalation by severity. Automation that knows when to stop.
Agentic & AI
Infrastructure
Security & Compliance
Featured Work
PilotCustomer Escalation Triage
A governed intake pipeline, not a demo chatbot
Problem
Customer success teams get escalations, bugs, billing fights, and churn signals by email. Manual triage is slow. Blind automation sends the wrong reply.
Approach
Each inbound email becomes one Slack thread. Workers classify severity and account context. Nothing customer-facing goes out without a human and an explicit reply marker.
Architecture
Shared state: one Slack thread per inbound email. Click a step to expand.
Gmail
Inbound trigger
Escalations, bugs, billing disputes, and churn signals arrive by email. Each message is one unit of work.
Orchestrator
Intake and classification
Pulls the email in, classifies it, posts a Slack summary, and writes structured fields to the table.
Slack + Table
Shared state
One Slack thread per email. The table holds fields agents and people update. All processing stays in the thread.
Severity Workers
Routing and scoring
Workers tag severity and route by account tier. Billing, churn, and escalations are flagged for human review.
Human review gate
Approval before external action
No customer-facing reply leaves without a person. High-risk categories never get an autonomous path.
__REPLY__ send
Governed outbound reply
External replies require an explicit marker. Unmarked messages stay internal.
Architecture walkthrough and sanitized flow demo available on request.
Orchestrator
Gmail to Slack and a table. Classifies issue type, posts a summary, stores thread ID, customer email, account tier, and category.
Severity and Routing Worker
Tags and escalates by severity and account tier. Flags billing, churn, and escalations for mandatory human review.
Python Priority Agent
Deterministic severity scoring from structured fields: account tier, issue category, escalation recency, and contract value.
Governed Reply Sender
External replies require an explicit __REPLY__ marker. Unmarked thread messages stay internal.
Governance safeguards
- Human review required before any external customer reply
- Billing and churn signals always escalated with no autonomous resolution
- Explicit reply marker prevents accidental customer-facing sends
- Severity matrix with documented escalation paths by tier and issue type
Education & Credentials
-
Agentic Specialist Path CompletingHireHuman
Reliable agentic workflows: evaluation, orchestration, Python agents, launch readiness
-
Professional certification CompletedCoursera
Details available on request
FAQ
Straight answers about the work.
What is Vadrel?
Vadrel is the firm Zac uses for client work. This site is him. The firm puts AI into work companies already do: build it, teach the team, or both.
Do you still do cybersecurity?
No. That chapter is closed. Nomad Cybersecurity Group is in the past. New work is AI operations and teaching, not security auditing.
How do you work with a company?
Two ways. I build the thing and leave it running. Or I sit with your people and teach them to run it. Most clients want some of both. You do not get a deck.
Are you taking work?
Yes, a small number of companies at a time, through Vadrel. Email or LinkedIn.
Research Zac with AI
Click a platform to read this site and LinkedIn, then get a plain summary of who Zac is and what he does. Client work goes through Vadrel.
If a platform does not auto fill, paste the copied prompt after it opens. Machine readable context also lives at /llms.txt and /llms-full.txt.
Preview research prompt
Open to load prompt preview.
Let's put AI into the work you already do
Client work goes through Vadrel. This page is me. Email or LinkedIn.
Resume available on request via email or LinkedIn.