Software built by a cyber mind · Lisbon, Portugal
Invoices read into your systems, routine email answered, estimates drafted — built so you know how often it’s wrong, what it can see, and what it costs each month.
Starts with a two-week audit · €4–6K fixed
Andrew Kovalenko Principal engineer, twenty years of shipped software. One person — the one who scopes it builds it.
Shipped at, ongoing for, or built by
Accounting, tax & AI
Streaming & OTT
Multi-brand CPG
Fixed scope, fixed price. You’re buying a defined outcome, not a block of someone’s calendar.
AI, put to work
Getting a model to do something impressive once is easy now. The engineering is everything around it: knowing how often it gets things wrong, keeping it away from data it shouldn’t see, connecting it to the tools your people already use, and knowing what it will cost when everyone uses it. That’s the part I do.
01
Before anything goes live, I build a test set from your real documents and cases and measure the error rate. You decide what’s good enough; what the system isn’t sure about goes to a person instead of being guessed. Arithmetic stays in ordinary code — the model reads, it doesn’t do the maths.
02
It gets the permissions of the person using it and nothing more, and anything it wants to change is checked again before it’s applied. Which provider processes your data, where, and whether it may be used for training is settled in writing before any real data is sent.
03
Cost per document, per request, per user — measured, and projected at ten times today’s volume. The cheapest model that passes the test, and hard monthly limits, so the bill never surprises you.
Pricing
Fixed scope, fixed price, never hourly — a defined deliverable with a date. The engineer who scopes the work is the one who builds it.
Start here
€4–6K fixed · 2 weeks
I walk through the work with the people who do it, pick the two or three places AI pays for itself, say where an off-the-shelf tool is enough, and test the top one on 50 of your real documents — error rate and monthly cost measured. Already have a prototype? The same two weeks check it for accuracy, security and cost instead.
Build
from €12K fixed · 4–6 weeks
One defined piece — an AI workflow, an app, an integration — connected to the tools you already use: email, Google Workspace or Microsoft 365, your CRM or accounting system. With tests, monitoring, spending limits and a handover to your team.
After launch
€1.5K per month
Models change every few months, and so do your documents. Monitoring, model updates, a monthly accuracy and cost report, and small fixes. Recommended for any team without its own engineers.
Procurement & ops
Selected work
A Portuguese tax and bookkeeping product where AI reads receipts, invoices and bank statements, categorises transactions and answers questions from the user’s own ledger — while a tax engine does the arithmetic. Every AI call is metered, each account has a monthly AI budget, and the assistant reaches only the signed-in user’s data. The same three rules I bring to your business, running in production.
From engineers I’ve shipped with
We worked together on the Locast.org project. Andrew brings an outstanding mix of strategic thinking, technical skills, and hands-on problem-solving — and a consistent high standard of quality.
Five years working with Andrew. Deep understanding of the trade-offs in high-traffic applications, project infrastructure, and system design. His AWS expertise is admirable.
Deep technical expertise plus exceptional communication. Andrew brings thoughtful architecture, reliable delivery, and the ability to turn ambiguous requirements into pragmatic solutions.
One of the most effective and reliable people I've met in my career. Strong technical skills, and the ability to set priorities and keep difficult projects moving.
Direct client references available on request.
I spent most of two decades building streaming platforms. This year the last of that work moved to SaaS — and I noticed my hands had already moved somewhere else.
What one person with two decades of shipping can now own — and what hasn't changed.
After twenty years inside teams, running a studio solo is a choice — not a stage. The reasons hold up better than I expected.
Frequently asked
Contact
One inbox, no ticket queues, no BD layer. Tell me what your team spends too much time on, or what your AI prototype can’t do yet. If two decades of engineering plus modern AI can help, I’ll say so. If not, I’ll say that too. Most engagements start with the two-week audit — a small, fixed first step for both of us.
— Andrew Kovalenko, CYBIND
