Nordic DevhousePublic Lite

Nordic Devhouse / public portfolio / five products

Web data,
made usable.

Six products from one software company, built on a reusable system for turning search-driven websites into structured, normalized, enriched, and productized software.

01 / Products

Five expressions of one system.

02 / Scraping engine

The system behind every product.

An engine that algorithmically generates, validates, and repairs scraping code for any search-driven website.

The input is a website and a search objective—not a manually written scraper. The engine explores how that site searches, filters, paginates, and presents results. From those observations it builds a site model, generates executable scraping code, and validates the implementation against real results before data enters the product pipeline.

Traditional scraping

One developer writes and maintains one scraper for one website.

This system

The algorithm discovers the search behavior and generates the site-specific implementation.

Website + search objectiveAlgorithmic explorationSite modelGenerated scraperAutomated validationUsable data

How it works

The generated implementation is specific to the target website. The method that produces it is reusable.

  1. 01

    Define the target

    A website, search objective, desired entities, fields, coverage, and freshness requirements establish what successful extraction means.

  2. 02

    Explore the search surface

    The engine tests queries and interactions to identify request patterns, filters, pagination, infinite scroll, result cards, and detail-page paths.

  3. 03

    Infer the site model

    Observed behavior becomes a structured map of inputs, navigation states, result boundaries, field locations, and relationships between pages.

  4. 04

    Generate scraping code

    The model is compiled into executable, site-specific code for requests, traversal, field extraction, deduplication, retries, and error handling.

  5. 05

    Validate real results

    Automated tests run the generated scraper against the live site, checking completeness, field quality, duplicates, coverage, and expected result counts.

  6. 06

    Extract at search scale

    Validated code mines the full search-result space—not only individual pages—while preserving source references and collection timestamps.

  7. 07

    Normalize and enrich

    Site-specific values are transformed into a consistent schema, resolved across sources, and enriched with provenance, freshness, confidence, and derived attributes.

  8. 08

    Monitor, repair, and deliver

    Structure changes trigger validation failures and a new exploration-and-generation cycle. Healthy data is delivered as datasets, APIs, monitoring, or product workflows.

What the generator produces

Not a generic crawler. A working implementation for the target site.

  • Search request generation
  • Filter and query logic
  • Pagination and infinite scroll
  • Result-card and detail-page traversal
  • Field extraction rules
  • Deduplication and identity logic
  • Retries and error handling
  • Validation tests
  • Structure-change detection
  • Regeneration and repair signals

03 / LLM context

Give an LLM the project, not just a screenshot.

Copy one automatically maintained, machine-oriented context package into ChatGPT or another LLM. It includes the portfolio map, shared scraping engine, product maturity, claim boundaries, and useful reasoning guidance.

Open plain text ↗

Generated from the public product registry; no private Full content is included.

Company: Nordic DevhouseProducts: Scraping Operations, Price Compare, Entity Intelligence, Dataset API, Custom Software Studio, Data Monetization, Outbound Lead GenerationCore: website + search objective → site model → generated scraper → validation → extraction → normalization + enrichment → deliveryScope: generated public context; private Full archive excluded