Software & data engineering · Est. Sweden

We build the systems that make data worth trusting.

We design warehouses and pipelines, build the applications on top of them, and stay long enough to see whether the numbers still hold six months later.

Start a conversation
WarehousesModelled & tested
ApplicationsGrounded in real data
PlatformsBuilt, run & licensed
InfrastructureAWS · Kubernetes
01

Three things we do, properly.

01 — Data platforms

From raw source to a number you can defend

Warehouse modelling from ingestion to the tables a business actually reports on. Layered transformations, tests that fail loudly, and lineage you can follow when a figure looks wrong.

  • dbt · medallion modelling
  • Snowflake · PostgreSQL
  • Ingestion & reconciliation
  • Data quality gates
02 — AI applications

Assistants grounded in your data, not a model's imagination

Retrieval over real records, tool use with guardrails, and evaluation harnesses that catch regressions before your users do.

  • Retrieval-grounded assistants
  • Tool use & agent workflows
  • Evaluation & regression suites
  • Cost and latency budgets
03 — Platforms

Some systems we do not hand over

We build the platform, operate it, and license it to the business that trades on it — so the operator runs a business instead of an engineering department.

  • TypeScript · Next.js
  • AWS · Kubernetes
  • Search & catalogue at scale
  • iOS & Android delivery
02

Most data work fails quietly.

A pipeline keeps running. A dashboard keeps rendering. And the number on it stopped being true three weeks ago — with no alert, because nothing crashed.

So we build for the quiet failure first. Every transformation carries tests. Every derived table can be traced back to the row that produced it. When a figure moves, the question why did it move has an answer that takes minutes, not an afternoon of spreadsheet archaeology.

The same discipline applies to the applications on top. A feature is not finished when it renders — it is finished when we have exercised it end to end against real data, watched it under load, and can point at the measurement that says it works.

We prefer small teams and direct contact. There is no account layer between you and the person writing the code.

01Correctness is measured, not asserted.
02Every derived number keeps its lineage.
03Ship in slices; verify each slice in production.
04The people who build it, run it.
03

Selected work.

Automotive commerce · Built, run & licensed

Otiva

An automotive spare-parts commerce platform: hundreds of thousands of parts reconciled from multiple wholesale suppliers, matched to vehicles down to engine level, and served through a public storefront with native iOS and Android clients.

We built the platform and we run it — supplier ingestion and deduplication, the warehouse models behind the catalogue, fitment matching, the storefront, the mobile clients, and the infrastructure underneath. The platform is licensed to its operator, who handles the commercial side: stock, pricing, orders and customers.

otiva.com.tr
SUPPLIER FEEDS RECONCILE SURFACES CATALOGUE + FITMENT Wholesaler AWholesaler B Wholesaler CWholesaler D StorefrontiOS app Android app
04

Company details.

Legal name
DataForte AB
Company type
Aktiebolag — Swedish limited company
Registration number
559523-6323
Registered office
Tant Gröns Väg 54, 147 60 Uttran
Stockholms län, Sweden
Enquiries
We take on a few engagements at a time. Tell us what you are building and what is currently in your way — a short description of the system and the problem is enough to start.