Internal systems · Digital products · Process automation

What I can own end to end

One accountable owner for product, architecture, data and production — no handoffs between vendors. For a bounded system that has to reach real operation, not a presentation.

Discuss a problem About

18+ years in operations and infrastructure2 products in 2026: production and acceptance85+ end-to-end tests in CI

What you can hand over

A team can do this — but most of the time goes into handoffs between roles. When the system’s boundaries are clear, one owner with the full context decides faster and keeps the architecture whole.

Custom internal software

Your operations live in spreadsheets and chats. Build the system around your process.

Data diverges between branches and people, “only your own” access is impossible, reports are assembled by hand, off-the-shelf products don’t fit your roles.

What you get

A system your people use every day: scoped roles, one data model, an audit log
Reports and dashboards, import and export, backup and restore
Deployed in your network or in the cloud; role manuals for handover

I need an internal system →

People and branches
Roles and permissions
One data model
Reports and audit

9roles, ~140 permissions
26tables
85+e2e tests

More: why me and how the work runs

Why me. In 2026 I designed and built exactly this for a network of sports schools: head office and branches, 9 roles, ~140 atomic permissions, athlete transfer between organisations, audit log, Excel, 85+ end-to-end tests, offline installation; acceptance with the customer against test protocols. Before that — 18+ years of designing roles, permissions and processes for real organisations.

How the work runs. A short review of requirements and the data model with your people → architecture and a staged plan → build with regular demos → acceptance against a protocol → handover with documentation. Then support as agreed.

Digital product to production

You have the product idea. You need someone to take it to production.

There’s an idea or a prototype, but no one to carry it to a product people pay for and that doesn’t fall over at night: accounts, scheduling, bots, subscription, infrastructure.

What you get

A product in production on your domain: user and admin sides, integrations
Reminders and notifications, subscription or another monetisation model
Infrastructure with TLS and backups — a foundation to iterate on, not rewrite

Discuss a product →

Idea
Product model
Build
Production
Iteration

31pages, 3 bots
liveSoloHairCut, since 2026
1 dayMVP on an existing architecture

More: why me and how the work runs

Why me. SoloHairCut — an online booking platform for stylists in Serbia, in production since 2026: 31 pages, a dashboard with three scheduling modes, three Telegram bots, reminders, a digest, subscription. Belgrade Stay — direct booking for hostels: an MVP in one working day on the previous product’s architecture, 38 of 38 UI tests. The foundation is reused — the second product comes together faster.

How the work runs. A review of the product and the boundaries of version one → architecture on a proven foundation → a first version with real users → iterations driven by data. Infrastructure and operations from day one.

Business process automation

A process that runs on people — automate it, including the steps that need text and speech.

The process eats staff hours and makes mistakes; data from external systems is merged by hand; some steps need text understood — correspondence, documents, speech.

What you get

Integrations with external systems and APIs, bots and notifications, import and reconciliation, scheduled reports
An LLM layer with a human check — only where text or speech has to be understood
A before/after metric in hours, errors or speed, agreed up front

Automate a process →

Manual process
API · rules · LLM where needed
Automated workflow
Measured result

24 / 2 hreminders + digest
208vehicles from a GPS API (MVP)
offlinevoice translator, swappable AI layers

More: why me and how the work runs

Why me. In my own products: three Telegram bots with reminders 24 and 2 hours ahead and a morning digest; analytics of training and medical data; a voice translator with swappable recognition, translation and voice layers, including an offline mode; monitoring of a 208-vehicle fleet from an external GPS API (MVP). Historically — automation of deployment, monitoring, reporting and backups in two companies, where it changed the economics.

How the work runs. Map the process and measure “as is” → the smallest change with the largest effect → run alongside the current process → compare against the metric → switch over.

Prototype → production · additional

The prototype works. Now it needs production.

The first version was built fast, and it can’t be shown to customers or put into operation as it is.

What you get

A review of architecture and data model, roles and permissions, tests, CI
Infrastructure with TLS and backups, a restore procedure, documentation
The “fix or rebuild on the same foundation” decision made on data

Take it to production →

Prototype
Tests · roles · infra · backup
Production

v1 → v2both of my large systems
400+users — the recovery the vendor considered impossible

More: why me and how the work runs

Why me. Both of my large systems went through v1 → v2; in the flagship that meant a reworked data model and a separate end-to-end test suite in CI. Before the solo period — two data centres with no single point of failure and backups with regular restore drills.

How the work runs. Audit and plan → staged hardening, starting with whatever could take production down → handover with documentation. Requires access to code and data and the freedom to change the architecture.

Discuss a problem

How the work runs

Understanding the problemwhat we do, what we don’t, how we measure
Architecture and plandata model, roles, stages
Build with demosa working part on staging, not a report
Acceptanceagainst an agreed protocol
Production and handoverbackups, documentation; the system is yours

We agree on results and stages, not hours. Architecture, data, the access model, infrastructure, backups, review and production are my responsibility; a significant share of code, tests and documentation is written by AI agents under project rules — faster, but it doesn’t change who answers for the result.

The two projects I refer to

SoloHairCut

Online booking for independent stylists, Serbia. My own product: product, architecture, data, backend, frontend, integrations, infrastructure, operations.

solohaircutstudio.com, in production since 2026

Three Telegram botsclient with a Mini App, stylist, admin
Stylist dashboard20+ endpoints, three scheduling modes, clients, subscription
Operations8 services in Docker Compose, automatic TLS, scheduled reminders and digest

31pages
16tables
8services

System for a network of sports schools

Training and medical monitoring: head office and branches. Customer under NDA; requirements, test protocols and acceptance — with the customer on staging.

The system’s role model: 9 roles, scope own / organisation / all
Role	Groups	Athletes	Diary	Medical	Transfers	Morning check	Wellbeing
System administrator
System moderator
Branch moderator
Director
Head coach
Senior coach
Group coach
Doctor
Athlete

My ownershipspecification, architecture, data, role model, backend, frontend, CI, e2e, offline installer, handover package
Data and controlmultiple organisations with athlete transfer, audit log, Excel, backup and restore
TechnologyReact 18 · TypeScript · FastAPI · PostgreSQL 16 · Playwright · Docker · GitLab CI

9roles
26tables
85end-to-end scenarios

Also: a voice translator with an offline mode (MVP) · Belgrade Stay — hostel booking (MVP in a day) · a task tracker for service companies (v1, 167 tests) · monitoring of 208 vehicles (MVP) · 400+ · ~96 h email recovery — The story →

When I’m not the right fit

You only need a landing page or a simple website.
You need development billed by the hour.
The problem needs a large delivery team from day one — then see the organisations page.

Have a problem that needs to become a working system?

Describe it in 2–3 sentences: what the system or process is, who will use it, what hurts. I’ll reply with how I would approach it; if it’s not my kind of work, I’ll say so straight away.

Discuss a problem About

Emailsaltykov.nikita@hotmail.com LinkedInlinkedin.com/in/nikita-saltykov

Working in a company with its own team? What I offer organisations.