About
I’ve seen the AI-vs-reality gap from every seat.
Founder, operator, analyst, engineer. That sequence is rare, and it’s the whole point — I know where AI projects break because I’ve stood in each place where they break. Here’s the path.
The arc
An unfair advantage, built one seat at a time.
- 01 · Engineer
Avionics engineer by training
Moscow State Technical University of Civil Aviation — M.E.
The foundation: systems thinking from a field where a missed edge case is not a rounding error. I learned to reason about how parts fail together, not one at a time.
- 02 · Founder
Founder & CEO — 17 years
Effective Systems Ltd. (Moscow)
Built from zero to $1M+ in revenue: 17 product lines, 260+ SKUs, 1,000+ installations, clients including national energy majors. I wrote the automation myself — MySQL, Python, C#/Modbus, and early machine learning for energy use. I have signed the front of the paycheck, which changes how you think about ROI.
- 03 · Analyst
Reskilled, deliberately
Hult International Business School — MBA + MS Business Analytics
Not a pivot — a deepening. GPA 3.87, #1 on the DataCamp leaderboard, and won the Hult Global Hackathon building a neural network from scratch. I went back to school to put rigor under the instinct.
- 04 · Operator
Operator-analyst in the field
HomeWorks Energy — 3.75 years
Where the theory met the schedule. 71+ hours a week clawed back with PDF-parsing automation — plus ~10 more from auto-creating Salesforce records, and field crews who stopped making second visits and redoing paperwork once they had an AI assistant on hand. A Salesforce ↔ ServiceTitan ↔ NetSuite integration on AWS Lambda/EC2. 11 internal tools and ~60 AWS Lambda functions — a whole automation backbone for one client. $120K in HVAC revenue growth. This is the seat where you learn the data is always messier than the deck implied.
- 05 · Engineer, again
AI engineer & architect
General Informatics → Forms On Fire / Appenate
RAG tools, custom bots, and AWS + n8n automation for SMBs at General Informatics. Then a constraint-programming scheduling engine and an internal MCP server that gives AI agents governed, audited access to enterprise data. The AI work, grounded in everything that came before it.
How I work
A bias toward removing complexity.
Three vendors mean three places a process can break — and no one owns all three. I’d rather build one system that works and stays owned than wire together a stack nobody can debug.
I’ll tell you when you don’t need AI. Often the honest answer is a deterministic automation, a data contract, or a process change — and saying so is worth more than selling you a model.
Every engagement is measured against one question: did it change a decision, or just produce a dashboard? ROI shows up when the work removes a step — not when it removes a name from payroll.
Want the long version?
The fastest way to understand how I work is to tell me what’s broken. If you don’t need AI, I’ll tell you that too.
Open to contract AI-automation engagements across the US — on-site, hybrid, or remote.