I help people meet their inner scientist.
Forward-deployed means I work inside the client's team: their systems, their data, their weekly call. Remote by default, on site when it pays. I build small data tools on the platform the client already pays for, then teach their team to run them.
- Who I work with
- Manufacturers that sell through distributors. The dealers and distributors themselves. Employee-owned companies and the nonprofits that serve them.
- What I build
- Reorder flags, supplier scorecards, dealer portals. Tools that find where the money leaks and stop it.
- How I work
- On the client's weekly call, with the people who use the tool. We set up an experiment and learn from it together. The teaching is the part I like most.
- Proof
- One weeks-of-supply tool, $5.5M in orders above baseline over four years. The managers still open it.
Builds run through Deanalytics, my firm. I'm also open to forward-deployed engineer (FDE), solutions engineer, and data engineer roles with AI teams in San Francisco, especially where the customer is employee-owned or mission-driven. Whatever I take on, I finish and hand over.
Who's my audience?
That is the first question I learned to ask at Bard, and it is still the first one I ask on a job. You have to meet people where they are, intellectually and sometimes physically, to help them solve their problems. Forward-deployed is the engineering word for the same idea. Some of what I've built, I built at a desk inside the client's building. Most of it, on a weekly call with the people who use it. The call is the desk.
Find out what they need
The brief says "dashboard." The planner needs to know which parts to buy this week. I find that out before I write a line.
A working tool in weeks
On the platform the client already pays for, in a form their team can open and change. Small and plain, and it's in production.
Stay until it runs without me
I stay on the weekly call and fix what's wrong. I tell the owners and executives what the data can and can't say, so they can decide what to spend and what to do first. Then I teach the person who keeps it running after I leave.
The ask was a dashboard: show inventory across the distributors. Before I built it, I sat with the regional sales managers who would use it and asked what they did with the number. They called distributors and suggested reorders, from memory and a spreadsheet.
So the tool became a reorder flag: weeks of supply by SKU, by distributor, refreshed from the distributors' own sell-through. A manager opens it, sees who is about to run out, and makes the call. Small, and in production on the platform they already paid for.
The first four months measured $2M in orders that followed its flags. Four years on, that is $21M, and $5.5M of it is above what the old pattern would have produced. The managers still open it. I'm still on the weekly call.
Forward-deployed data work, 2010 to now
Deanalytics, LLC · Founder
A consulting firm I run from San Francisco, full time since 2023. The current flagship is a four-year engagement with a legacy manufacturer that sells through distributors: weeks-of-supply by SKU across the channel, a weekly buy list the planner opens every Monday, supplier scorecards, and a dealer portal now in pilot. The weeks-of-supply tool alone has $5.5M of orders above baseline against its name. Also: a sales performance scorecard and supplies analytics across tens of thousands of printers for an office-technology dealer; and open research on employee-owned companies. On the AI side: a cascade classifier in production that sorts employee-owned companies by sector, built on the Anthropic SDK and tested on 526 hand-labeled cases across 21 sectors. Routing on a binary flag instead of a confidence threshold cut false negatives by 65 points.
RXA, LLC · Senior Director of Data Strategy
Directed a 12-person applied-AI and BI practice. Delivered 75+ engagements in biomedical, rapid manufacturing, agriculture, and healthcare, including five years of KPI and profit reporting for a produce grower.
O.F. Mossberg & Sons · Data Science Manager
Built the analytics function at a century-old manufacturer: five reporting tools into one cloud warehouse, $115K a year in net savings, and a lean coordination system that cut waste 24% and excess inventory 25%.
Markertek Video Supply · Data Analyst & BI Manager Employee-owner
Started in sales, ended running BI. Built A/B testing on the customer base; average order value up 34%, marketing spend down 33%. The company became 100% employee-owned in 2016, which is where my interest in ESOPs comes from.
Earlier work
Solar installer and the trades, after college. I still like wiring things.
A manager and two clients, in their own words.
“Aaron was able to accomplish the goal of building a data driven culture to grow with the business by implementing the Domo platform. In addition he built a tool that delivered self service capabilities so the business was not 100% dependent on IT.”
Wes BeighleyCIO, O.F. Mossberg & Sons. Managed me, 2018–19.
“Aaron is personally invested in his clients’ success — he’s not only a Domo expert but a great teacher. The level of communication and empathy is unique, not just in a vendor relationship, but in business generally.”
Todd LydaSenior Manager, Market Insights & Growth Enablement, RJ Young. Client.
“Aaron understands employee ownership deeply, as working for an employee-owned company changed the trajectory of his own life.”
Katie LehmkuhlMichigan Center for Employee Ownership. Client.
The rest of me
After the pandemic I wanted my life back: the autonomy to run my own business, to be in charge of my own health, and to catch up on the exploring I'd put off. The chance came and I took it.
I worked from the road for two and a half years, through Europe, Central America, and the Caribbean. It was special. It also taught me more about running a small company than any year at a desk.
I settled in San Francisco because I wanted a place where creativity, expression, technology, and possibility coincide. I wanted to be near the arts. I'm a new musician, piano and voice. Learning music as an adult has been a great gift. I like the challenge.
I've lived with ulcerative colitis, an autoimmune condition, for fourteen years. I wrote a travel blog called Have Guts, Will Travel about getting on planes anyway. The videos are on YouTube, and one post is still up.
The .mba in the address is real: SUNY New Paltz, 2017. Before that, history and Middle Eastern studies at Bard, a semester at West Point, and Arabic at a language institute in Fes. A history degree turns out to be good training for data work. Adam Weinberg, Denison's president, calls the liberal arts "what gives the data sciences its heart, its soul, and its greatness." I think he's right.
I like making things
Wood, steel, wire, thread, plastic. A few years of weekend projects, plus a 115-device home automation setup I ran for three years. It's why I'm comfortable with hardware. I trust a tool more once I've seen it break.
Also: a waterjet-cut toilet vacancy sign, a headset battery stand, and more embedded code than the photos suggest.
Mostly about employee ownership and data
The long versions live on the Deanalytics research page. The short list:
- ESOP × AI Exposure
- Aggregate ESOP Distributions 2009–2024
- The Retirement Horizon Problem
- Do Employee-Owned Companies Say So?
- Data Literacy in ESOPs
- How to Launch a Data Committee and How to Start a Data Dictionary
- "Aaron Dean: The Game-Changer in ESOP Data"
- NCEO Annual Conference and Fall Forum, 2025
- The ESOP Almanac, ESOP Scanner, and ESOP Jobs
What I reach for
- Data
- SQL first. Python for pipelines and models. Cloud warehousing and ETL on messy operational and government data.
- Platforms
- Domo (Major and Partner certified), Power BI, Tableau, HubSpot. Supabase and Cloudflare on my own builds. I build on whatever the client already runs. If it has an MCP server, I can build on it.
- AI in production
- Anthropic SDK deployments with evaluation sets, budget gates, and a human in the loop. Python and JavaScript, shipped with AI coding tools.
- Hardware
- ESP32, Zigbee, Arduino, soil sensors driving a pump.
Say hello
I'm in San Francisco and can be on site three days a week. For a build, tell me what your team is stuck on. For a role, send me the posting. For coffee, say so.