Roger Henley

VP Engineering & Technology, Whiskey House · Director of Engineering, Bardstown Bourbon

Industrial automation, plant operations, applied AI.

Thirty years running regulated 24/7 industrial systems, the last eleven in distilled spirits and the last eight commissioning two startup distilleries from concrete pour to steady-state, in the kind of shifts you don’t brag about. I am looking for a senior engineering or technology leadership role, including Principal and Senior Principal positions. Open to relocation for the right role and team.

JD review
What is this?

Paste a job description and it is read requirement by requirement against my records. Each requirement comes back met, partial or unmet, with the evidence behind the verdict so you can check the reasoning rather than take a number on trust.

When the fit comes back strong or better it also writes a two-page résumé for that specific posting and returns it as a PDF. Every line of it is drawn from my actual records, with the sources recorded. Nothing is invented.

There is a daily limit per account.

Sign in for jd review
Book a meeting
What is this?

Pick a length, 15, 30 or 45 minutes, then a day and a time. The slots come from my real calendar, so anything offered is genuinely free, and booking writes the meeting to it.

Choose video or phone. For video you pick Google Meet, Microsoft Teams or Zoom and host the room. For phone, give the number you will call from.

Both sides get an email with the details, and you can cancel your own booking from the same page. Times are shown in Eastern Time and follow daylight saving.

Sign in for book a meeting
Ask Roger
What is this?

An AI assistant that answers in my voice, from my own records, and shows you where each answer came from. Ask about my background, what I have built, or what I have written.

Answers take a few tens of seconds. Everything runs on one small server and it reads my actual records before it says anything. Questions I have answered myself come back instantly.

It will not discuss compensation, references, anything confidential to an employer, or my personal life. Those are better asked of me directly. Asked something my records do not cover, it says so rather than guessing.

Sign in for ask roger

Members only. Request access, or sign in. Hiring, and would rather just talk? Reach out directly.

The writing and the work photos and the podcast appearances are open; read them first. An account is for the reviewer: paste a posting and it reads it requirement by requirement against thirty years of manufacturing and applied-AI records, then writes a two-page résumé for that specific job. It is also how you book time with me.

in the field
30yrAcross power, mining, telecom, industrial automation, and applied AI.
distilled spirits
11yrSazerac, then eight years in two startups: Bardstown Bourbon Company and Whiskey House of Kentucky. Concrete pour to steady-state.
schedule
24/7Regulated, continuous production. The kind that pages you at 3 a.m.

why this site exists

The screen happens before the conversation.

Thirty years across oil and gas, power, mining, telecom, manufacturing, industrial automation and applied AI does not compress into a document somebody scans for twelve keywords. The terms that matter are usually equivalent rather than identical. A historian is a time-series database. A control narrative is a specification. Model predictive control is applied ML, optimisation under constraints.

Someone who does not recognise the equivalence screens the candidate out, and it happens before anyone has spoken. That is what makes it expensive: there is no conversation in which to correct it.

So this site does not hand you a résumé and hope. Paste the posting instead. The reviewer breaks it into requirements, goes looking through my records for evidence of each one, and returns met, partial or unmet with the passage behind every verdict. The score is arithmetic on those verdicts, computed in code, not a number a model felt like producing.

It reads your requirements and looks for evidence. It does not read my document and look for keywords. That inversion is the whole argument.

It is slow, it runs on one small server, and it disagrees with me often enough to be worth measuring. All three numbers are published, including which direction the disagreements run.

Top bubble tray of a continuous distillation column, viewed through the sight glass, rows of vapor caps against a wet copper wash.
column top bubble tray · post-run

The abstract stuff has to survive a copper still, a fermentation cycle, and a plant manager who’s been up for eighteen hours.

On why control theory and production reality have to meet somewhere.

the practice

Five areas, usually mixed together in one role.

Naming them separately makes it easier to see where I would fit on your team. Every real project pulls from at least three.

recent writing

What I’ve been thinking about.

All writing →

on other people’s shows

Two Guys and a PLC.

Two conversations about controls, plant systems and where applied AI actually earns its place on a production floor.

Opens on Spotify. Nothing from Spotify loads on this page until you choose to go.

Roger at a University of Kentucky podium giving a talk on enabling citizen developers with low-code tooling and BI.
University of Kentucky · talk on citizen developers

alongside the work

Talks, notes, and open source.

I write about the parts of digital transformation that don’t survive the vendor slides, how a control room actually adopts a new tool, how “citizen developer” enablement works when the citizens have wrenches on their belts, why the plant historian is a distributed database whether IT knows it or not.

Frameworks like MDEMG and Forge live on GitHub, free for use.

github.com/reh3376LinkedIn

A dragline excavator at Hobet Mining at dusk, its boom silhouetted against the sky over a stripped bench.

hobet mining · west virginia

Thirty years of on-the-floor places.

Coal draglines, telecom central offices, high-voltage substations, SCADA rooms, and, for the last eleven, distillery fermentation floors. The photograph is one of a long list; the rest are in the gallery.

A multi-monitor engineering workstation in a home office. A goldendoodle sits underfoot.
home workshop · the dog is not for scale

when the shift ends

A working room, a family, a dog with strong opinions.

Long hours in industrial jobs are only sustainable if the rest of a life is real. The workshop above is where the R&D happens between shifts, frameworks written, models trained, papers read, and the occasional deploy at midnight. The goldendoodle supervises.

A Whiskey House of Kentucky team meeting on the production floor during commissioning.
Whiskey House of Kentucky · team, mid-commissioning

built with a team

None of this is a solo act.

Distillery startups are built with operators, control engineers, instrument techs, safety officers, IT, quality, and a couple of contractors you learn to trust. The best days on this job are the ones where the team fixes something the org didn’t know it had, and no one takes the credit.

open source

Public repositories.

The frameworks and tools I work in the open. If one lines up with something you’re building, you can request contributor access below and I’ll add you.

Request contributor accessSee all on GitHub

what you can do here

Most of this is open. One thing is not.

The evidence is public, because you should be able to judge the work without asking anyone for anything. The reviewer needs an account, because running it spends real time on a real machine.

open to anyone

The writing
Every article in full, on control systems, manufacturing data and applied AI.
The work photos
A selection from the plant floor: the sites, the panels and the equipment behind the record.
How the reviewer works
Which model reads a posting, that it was never trained on the career it judges, and how far it currently disagrees with Roger. The numbers are live, including the ones he is failing.
A direct line
The contact form reaches him without an account.

with an account

Have a posting reviewed
Paste a job description. It is split into individual requirements and each one is judged against his record, with the evidence shown for every verdict and the score computed in code.
A résumé for that posting
When the fit is strong or better, a two-page résumé written for that specific posting, where every line has to be carried by a source. Delivered as a PDF and emailed to you.
Time on his calendar
Book a 15, 30 or 45 minute meeting against his real availability, shown in Eastern time. Confirm it in the app, or hold the slot and take the calendar file away with you.
The full gallery
The complete set of work photos rather than the public selection.
Your reviews, kept
Reviews reopen from your submissions list, so you can close the tab. Three reviews a day, because each one is close to an hour of work on one machine.

Access is approved by Roger himself, so a request is a short wait rather than an instant login.

Request accessRead how the reviewer works first

about this site

The site is itself a live work sample.

Go API and Python sidecar behind a Next.js front-end, all speaking Protobuf. Deployed with Docker Compose behind Caddy on a small Hetzner box. Source, spec, ADRs, and every decision that shaped it are public.

repo
reh3376/career-site
stack
go · python · next.js · postgres
deploy
docker · caddy · hetzner
contribute
open to contributors