Dallas Crilley
Now shippingForeman CT11:27:08 Dallas, TX
Applied AIForward deployedRevenue systems

I build the systems revenue and operations run on, and the AI inside them.

I spent ten years running the CRM, billing, and reporting stack at one company, and I know exactly where AI projects stall: the systems underneath never quite line up. My work starts with the operators, inside the stack a business already runs. I add agents where they earn their place, and I leave behind a system a person can trust: they can see what the AI did, check it, and stop it. A managed IT provider’s monthly billing has run this way for about 19 months.
CoHost AI Studio Stricter quality gate on CoHost transcripts
Throughline Retry-queue backoff in Throughline
focus applied AI · forward deployed · revenue systems
based Dallas–Fort Worth · CT · remote/DFW hybridAug 2026

Two skill stacks that don’t often overlap.

AI teams often hire engineers who have never owned a CRM pipeline or a billing reconciliation. Operations teams often hire generalists who have never shipped an evaluated agent system into production.

I sit between those lanes. Ten-plus years owning the systems that move work from a new lead to a paid invoice is the base layer: most recently as Head of Technology at Real News Public Relations, sole technical owner across six SaaS platforms and eight-plus production systems since 2018. Internal tools, applied AI, eval harnesses, and human review are the newer layer on top.

Those ten years were at a family-owned PR firm here in Dallas, where I’m still Head of Technology. Small company, no other engineers. When billing broke or a show didn’t air, it was mine to fix, and the people affected sat down the hall. I like that kind of ownership, and I like building tools the people around me actually use.

10+ yrs
Owning production systems across customer data, billing, reporting, and operations. Full history on hiring details.
202 accounts
One audit caught a billing bug before the next invoices went out.
7 systems
Public case studies, each written from the actual repository.
Recommended route for hiring managers

One shipped system. One applied-AI system. Then contact.

Meter shows the billing path in production. CoHost shows the quality gates around applied AI.

Selected systems

Things I’ve shipped. Each links to a case study.
All7AI automation2Tools and operations3Billing automation1Revenue systems1
Meter
Shipped
Billing automation
Intermedia usage, billed straight into ConnectWise.Moves Intermedia usage into ConnectWise invoices and removes the manual monthly reconciliation.
PythonConnectWiseIntermedia
~19 mo in production billing a real MSP
202 accounts in one audit incident (self-reported)
Read the case study
CoHost AI Studio
In progress
AI automation
Nothing publishes until it clears the gate.Checks audio, video, and transcripts before an episode goes live, and holds anything that misses the bar.
PythonTypeScripteval harnesshuman-in-the-loop
20+ pipeline steps
11 quality metrics
AI publishes only after checks pass
Read the case study
Throughline
Shipped
Revenue systems
One connector interface, every system in sync.Keeps six vendor tools in sync and puts each client’s information in one reliable place for operations and reporting.
PythonPostgreSQLOAuth 2.0retry queues
Read the case study
Live TV production automation
Shipped
Live operations
Audio-aware direction for live production.Helps a human operator direct live TV using audio cues and a clear studio state.
vMixPythonVB.NETstudio systems
Read the case study
EnrichCRM
Shipped
AI automation
CSV in, enriched contacts out, at a fraction of the cost.Turns a list of business emails into researched contact records, with the cost of each lookup visible.
TypeScriptNext.jsGeminiOpenAI
Read the case study
Foreman
Live
AI tools
Keeps your AI coding sessions alive and restarts the ones that hang.Watches a fleet of AI coding sessions, spots the stalls, and restarts only the session that hung.
TypeScriptBuntmuxCLI
Read the case study
Tether
In progress
Internal tools
Run your dev ops from your phone.Lets developers check on and approve coding-agent work from a phone.
SwiftmacOSCLI
Read the case study

Most of this work was private until August 2026, when I released two dozen repositories, prepared one at a time. The six below are the place to start. Each card links the file worth reading first.

Shipwright
An agent that takes an approved GitHub issue and returns a tested, reviewable pull request. Open the runner tests Repo
Holdfast38 tests
An append-only decision ledger with a human publish gate, enforced inside Postgres by triggers, a hash chain, and unique indexes. Open the attack tests Repo
Quorum173 tests
Human review as an API: agents submit work, real reviewers return a consensus verdict. Verify it yourself, offline Repo
Cinderwell342 tests
Disposable cloud dev servers that destroy themselves when a lease expires and write receipts proving it. Open the reaper tests Repo
Tenantwell30 tests
Postgres row-level-security multi-tenancy, tested by an adversarial suite that attempts cross-tenant reads. Open the isolation tests Repo
Reconciler
Finds billing discrepancies, proposes typed fixes, and requires a person to approve every invoice change. Open the verification workflow Repo

Want proof you can click before you clone? Seven live demos run real backends over synthetic data, no login. The proof page tracks all of it.

Recent work

Updated continuously
Continue for more context Architecture writing and how I work

Architecture writing

All posts
Adapting Cloudflare’s $1 Review Factory for RevOpsAgent systems · RevOps architecture
Cloudflare’s code-review factory, adapted to revenue data: coordinator, specialist fusion, risk-scaled compute, and the resilience work needed to make it usable as a demo. Read the article ~7 min read
The Four-Method Connector Contract, and Knowing When to StopThroughline · Connector design
A four-method connector contract that keeps six vendor integrations behind one engine-owned sync boundary, plus the restraint it took to keep the interface small. Read the article ~10 min read

How I work

What I own
The layer between AI and day-to-day operations: customer data, billing, reporting, internal tools, quality gates an AI output must clear before it ships, and the workflow logic operators actually depend on.
What I optimize for
Small surface, strong contracts, practical automation, and a clean handoff to the people who run the process after the engineering work ships.
Where I am strongest
When a process is manual, fragile, or hard to trust; when the data crosses systems that disagree; or when an AI workflow needs a quality bar, a review layer, and an operational owner.
Where I am not the right fit
Pure ML research, ground-up platform infrastructure, Kubernetes platform work, or roles that also require me to own visual design. I work at intermediate level on AWS and Google Cloud and ship with Docker, CI/CD, and managed platforms; I am not the person to build your cluster. I will tell you when a role is a stretch rather than overstate the fit.

Hiring for this kind of work?

Always glad to compare notes with people building applied-AI workflows, internal tools, and the business systems they run on. If you’re hiring for that work, email me. I read everything and reply within a day.

More ways to reach me

Recruiters and hiring managers can review role fit, level, comp target, location, and start date on the hiring details page.

[email protected] copy LinkedIn GitHub