Frederik Rybansky

AI InfrastructureAI AgentsBratislava, SK

Service 06

Custom Software Development

The non-AI half of the practice: APIs, integrations, internal tools and platforms that keep working for a decade. Most AI projects still need somebody who can build the boring parts properly.

Everything in the AI stack eventually touches an ordinary system: a database, a job queue, a webhook, an internal tool somebody uses forty times a day. Those systems deserve the same engineering discipline as the model layer, and they are usually where projects quietly fail.

This is the same practice that built BSS PRAHA platforms, the Audioline content system and BatteryCheck — Node.js, TypeScript, React, PostgreSQL, AWS and Docker.

Deliverables

APIs & backend services

TypeScript and Node.js services with typed contracts, versioning, rate limiting, idempotency and documentation your integration partners can use without asking you.

Web applications

React and Next.js front ends with real accessibility, sensible performance budgets and a design system your team can extend after I leave.

Integrations

ERP, CRM, payments, accounting, logistics, ticketing. The unglamorous plumbing work, done with retries, monitoring and reconciliation.

Internal tools

Admin panels, back-office workflows, data fixes, approval queues. The tools that remove the spreadsheet from your operations.

IoT & telemetry

Device fleets, ingestion pipelines, alert thresholds and reporting — as at BatteryCheck, where battery diagnostics run at fleet scale.

Team extension

Embedding alongside your in-house developers: architecture reviews, pairing, code review standards and handover documentation.

How I work

Understand the constraint

Not the feature list — the budget, the deadline, the systems you already run and the things you are not allowed to change.

Small, reviewable slices

Working software every week, on a branch you can read. No month-long architecture phase.

Boring where it matters

Postgres over a novel database, a monolith over microservices, well-understood patterns over clever ones — until the clever one is clearly required.

Operate it

Monitoring, alerts, runbooks and a deploy pipeline you can run without me. If it needs me to stay available, it is unfinished.

Outcome

  • A codebase your team can maintain without me.
  • Typed contracts at every boundary, so integrations fail loudly in tests, not in production.
  • Deployment you can run on a Tuesday afternoon.
  • Documentation that answers the questions you would otherwise ask me.
  • No framework that requires a migration in eighteen months.

Typical stack

TypeScript · Node.js · React · Next.js · PostgreSQL · Redis · AWS · Docker · Kubernetes · Terraform · GitHub Actions

Frequently asked questions

Do you only do AI now?

No — about half my work is ordinary software, and the best AI projects need both halves. If you need a CMS, an API or an internal tool built properly, that is a normal request and I will treat it like one.

Can you take over an existing codebase?

Yes. I start with a short review to establish what is real, then fix the highest-risk parts first — usually error handling, authentication, data integrity and whatever silently retries writes.

Do you work in your team's timezone?

Overlapping hours with CET, and I have worked with teams in the US and Asia by structuring around written updates and well-defined async decisions rather than meetings.

More services