Cristian Pogan

You started building with AI.
It got 70% of the way there.

I’m the technical partner who finishes it, deploys it, and keeps it running - web applications, sales systems and document automations that work on Monday morning, not just in a demo. Your servers, your code, costs you can see.

Book a 30-minute call

EST hours · Europe & US · Retainer or project

Sound familiar?

  • You built a working prototype with Claude Code. Then a quarter closed, something caught fire, and it has been sitting in a repo for two months.
  • The automation runs, but nobody trusts it enough to switch off the manual process behind it.
  • Your AI spend went up four times and you cannot tell which workflow is responsible.
  • You have been trying to hire an AI engineer since spring. You have interviewed eleven people.

None of these are prototype problems. They are operations problems, and they need someone who has run systems in production.

Scroll →

Where AI builds stall A progress bar from prototype to production. It starts at the seventy percent mark, where most builds stall - after the demo works, before anyone trusts it in production - then fills to one hundred percent and turns green: the remaining thirty percent is the operations work that gets finished. PROTOTYPE PRODUCTION 70% the stall - demo works, nobody trusts it 100% error handling · monitoring · cost tracking · docs
The last 30% is not more prototype. It is operations work.

Sales systems that generate meetings

Pipelines that find the right people, verify they are reachable, and put a conversation on your calendar - running on your infrastructure, not a vendor’s.

Lead sourcing & enrichment cascades · Email verification & deliverability · LinkedIn intent-signal detection · Automated campaigns · CRM configuration & data hygiene (HubSpot, Salesforce, Wealthbox)

40

Qualified meetings in 60 days for a US real estate firm - from an existing first-degree network and intent signals, not cold volume.

13,000

Verified leads sourced across 34 client accounts on a provider cascade that scaled without adding headcount.

Web applications your team runs on

Custom tools, portals and dashboards - built on your infrastructure, documented, and maintained after launch.

Custom web apps & client portals · Internal tools & dashboards · API integrations · Deployment, monitoring & maintenance

Document and workflow processing

The work that currently costs you a person: reading documents, moving data between systems, and chasing the exceptions.

OCR & structured extraction · ERP and accounting integration · Contract generation & automated signature · SOP knowledge graphs · Support routing · Agents running unattended on your infrastructure

400

Invoices processed end to end through OCR into the ERP for a European logistics operator. Data entry became exception handling.

See how these are built →

How the work runs

01

Discovery

one week · $2,500 fixed

I read what you have already built, map where the technical risk and the running cost actually sit, and hand you a phased plan with real numbers. Your side is about three hours. Yours to keep either way.

02

Build

scoped per phase

We take the plan in order of risk, not order of excitement. Everything is deployed to your infrastructure as it is built, documented as it is built, and visible to you daily.

03

Operate

from $4,999/month

Monitoring, cost tracking, maintenance and the next phase of work. This is where most automation projects fail, because most of them stop at build.

Engagement model Three sequential phases light up in order: Discovery, one week at 2,500 dollars fixed; then Build, scoped per phase; then Operate, from 4,999 dollars per month. Finally an accent line draws itself from Discovery back into Operate: the discovery fee is credited against the first month. 01  DISCOVERY one week $2,500 fixed 02  BUILD per phase scoped 03  OPERATE ongoing from $4,999/mo CREDITED TO MONTH 1
Discovery is credited against your first month if we start within 60 days.

You can see everything, every day

Every engagement runs in a dedicated Plane workspace - think Jira, but actually maintained.

Every decision from every meeting is written down. What is built. What is blocked and why. What we are doing about it. The plan measured against the objectives we agreed in discovery.

You open it whenever you want. You do not wait for a status call to find out where things stand, and you never have to ask me what I did last week.

Scroll →

The engagement board A kanban board with four columns: backlog, in progress, blocked, and done. When the board scrolls into the centre of the screen, one in-progress card - the invoice exception queue - slides across into done, turns green and gains a check, the done column makes room, the counts update, and a backlog card moves up into progress: how items get done. The blocked card states why it is blocked - ERP sandbox access pending client IT approval - and what is being done about it. Done cards are checked green, each deployed and documented. ENGAGEMENT WORKSPACE · PLANE written record · updated daily · open to you BACKLOG 3 2 IN PROGRESS 2 BLOCKED 1 DONE 3 4 new carrier format config change · phase 02 SOP knowledge graph phase 03 · scoped support routing rules chatwoot · phase 03 provider cascade order build · phase 02 decision record · from Mon call ERP posting API blocked: sandbox access why - client IT approval pending next - ticket #142, chased today invoice exception queue build · deployed as built deployed · documented mailbox ingestion deployed · documented rate-limit send guard deployed · walkthrough recorded credentials to your vault infisical · from day one you open it whenever you want - nothing waits for a status call every card → documented as it is built
The shape of every engagement board: what is built, what is blocked and why, decisions on record.

If you stop working with me, you lose nothing

Everything runs on your servers and your VMs. Credentials live in your own Infisical vault. Code sits in your GitHub with full commit history. Every integration is backed up daily.

Every feature is documented as it is built - written documentation plus a screen recording of me walking through it.

Your team can operate and modify what we build without me. That is deliberate. I do not want you dependent on my availability, and you should not accept an arrangement where you are.

There is no lock-in, because lock-in is not a business model. It is a liability you would be right to avoid.

Scroll →

What you own A boundary labelled your side of the line appears first. Inside it, three boxes light up: your servers with daily backups, your vault holding every credential, and your repo with full commit history. Lines draw from all three down into your team, which lights up green - it can operate the systems without me. Nothing sits outside the container. YOUR SIDE OF THE LINE - NOTHING LIVES ANYWHERE ELSE YOUR SERVERS daily backups YOUR VAULT every credential YOUR REPO full history YOUR TEAM docs + screen recordings, from day one
Every system, credential and line of code sits in your accounts. Your team can run it without me.

You will always know what it costs

Every API key is tracked against the workflow it serves. You get a weekly usage report showing spend per system, so cost is attributable to the thing that produced it.

AI costs rise quietly and rarely in proportion to the value they return. I learned waste reduction on a factory floor before I applied it to software, and the discipline is identical: measure it, attribute it, cut what is not earning.

Workflow Provider Week Per unit
Lead enrichmentApollo, Lusha$184.20$0.061 / lead
Email verificationDebounce$41.75$0.004 / email
Invoice extractionOCR + model$96.40$0.24 / invoice
Intent monitoringUnipile$58.00flat

Illustrative format. Figures are examples, not client data.

Tools I deploy and operate. The choice depends on your systems, not my preferences.

Who I am

Cristian Pogan

I founded and ran a computer-vision company for the recycling industry, leading seven senior engineers across Europe - hardware and software, in production, in factories.

I led development of RAG systems at the European Commission’s financial directorate, supporting assessment work on the €650 billion Recovery and Resilience Facility.

Before that, lean manufacturing at Dalli Production, a former Procter & Gamble plant.

I work with a small number of founders at a time as their technical partner. If I am full, I will tell you.

More about how I work →

Questions

Do you work with my existing developers?

Yes - most of my engagements involve an existing CTO, ops lead or project manager. I do not replace them and I do not need to own the roadmap. I take the systems work nobody has bandwidth for, work in your repo and your conventions, and your team sees every decision in the shared workspace. If your developers want to take something over, the documentation is already written.

What if I already started building this myself?

That is the best starting point there is. You have proved the idea is worth doing and you have made the hard product decisions. What usually needs work is not the logic - it is everything around it: error handling, credentials, monitoring, cost tracking, deployment, and the documentation that lets someone else maintain it. Discovery starts by reading what you built, not replacing it.

How fast can you start?

Discovery usually starts within two weeks of our first conversation. It is fixed scope and fixed price, so there is no long proposal cycle - we agree what we are examining and begin. Longer engagements start once discovery has told us both what we are actually dealing with.

What happens if you are unavailable?

Two versions of this question. If I am at capacity when you reach out, I will tell you directly rather than stretch across too many engagements, and I will give you a realistic date.

If you mean what happens to your systems while I am away: they keep running. Everything is deployed on your infrastructure with daily backups and full git history, credentials are in your own vault, and every feature has written and recorded documentation from the day it was built. Your team can operate what we have built without me.

What does it cost?

Discovery is $2,500, fixed. If we go on to work together, it comes off your first month - so if you hire me, discovery was effectively free. Retainers start at $4,999 per month. The credit applies if we start within 60 days.

What is included in discovery?

One 90-minute working session with you and your technical lead, read-only access to what exists, and your current AI tooling invoices. I spend the week reading the code and workflows. You get a written assessment: an architecture diagram of your current setup, ranked technical risks, your true running cost per workflow, and a phased plan with scope and price - including what your own team can handle without me. Then a 45-minute call to walk through it. About three hours of your time in total.

The playbook I bring to new projects

Fifteen technical insights from running these systems in production - what usually happens, why it keeps happening, and the system to run instead. Each one pairs the technical decision with the business metric it unlocks.

They are the defaults I apply to every new engagement, from your first technical hire to document pipelines, prospecting cascades and the founder context layer.

Read the AI insights I discovered so far →

Start with a conversation

Thirty minutes, no charge. Tell me what you have built and where it stopped. If I am not the right person, I will say so and point you somewhere better.

Calendar not loading? Open the booking page or email me.

EST hours · Typically booking 1–2 weeks out