Web scrapers, AI agents and backend services — scoped in writing, priced as a fixed number, delivered in days rather than quarters. Most projects never need a call.
We turned down the long service list on purpose. These are the areas where we can quote a price in an hour and ship something that survives contact with production.
Public websites, marketplaces and directories turned into clean, deduplicated datasets. Pagination, session handling, JavaScript-rendered pages and rate-limit politeness are handled properly rather than patched around. Output as CSV, JSON, Google Sheets or straight into your database.
Systems that do the repetitive work: lead discovery and enrichment, document processing, drafting and sending outreach, retrieval over your own files. Built with a local-model fallback where it matters, so a cloud API outage doesn't stop the pipeline.
Services, integrations and internal tools built security-first — authentication and authorisation, tenant isolation, presigned uploads, queueing, and deployment that doesn't fall over. Documented and handed across so you are never dependent on us.
The unglamorous half: cleaning, normalisation, deduplication, enrichment and recurring reporting. Set up once, scheduled, and monitored — so nobody on your team opens a spreadsheet on a Monday morning again.
Every extraction job ships with normalised fields, parsed prices and currencies, deduplication keys, and both CSV and JSON exports. If you want it in your own database instead, we write it there directly.
{
"scraped_at": "2026-09-07T18:04:11Z",
"source": "catalogue.example.com",
"pages": 42,
"rows": 1174,
"duplicates_removed": 63,
"records": [
{
"product_id": "SKU-88213",
"name": "Aluminium Roof Rail",
"price": 129.95,
"currency": "EUR",
"in_stock": true,
"image_url": "https://..."
}
]
}
// also written to output/products.csv
A sample of public and in-house builds. Client work under agreement is not listed here — we will talk through comparable projects on request.
Discovers businesses across LinkedIn and Google Maps, extracts contact details from their websites, generates a personalised message per contact, sends over SMTP and logs every send so nobody is contacted twice. Runs unattended with cooldown logic, and falls back to a local model when cloud APIs are unavailable.
A production-style catalogue scraper handling pagination, price and currency parsing, stock status and deduplication by product ID, exporting to CSV and JSON. Headless by default with configurable request delays, so it can run on a schedule from a server without getting the client blocked.
A Chrome extension that reads the live DOM and carries out instructions written in plain English — navigating, clicking, typing and extracting text into a memory scratchpad it can reuse across tabs. Provider-agnostic across Gemini, GPT-4o, Groq and local Ollama models, with bring-your-own-key so running costs stay with the user.
Describe the site or the task and the fields you need. Two minutes in a form — no call, no discovery workshop.
Within a few hours you get a flat price, a delivery date and what is explicitly out of scope. If it isn't a fit, we say so.
You see real output early — before the full run — so direction is corrected while it is still cheap to correct.
Final data, plus the script or workflow and its documentation if you want to keep running it yourself.
A small team by choice. The people below are the people who do the work — their repositories are public, so you can check the standard before you commit a budget.
Backend services, agent architectures and data extraction. Production work across CRM, KYC and fintech systems — authentication, tenant isolation and API security. BS in Artificial Intelligence.
GitHubMachine learning, exploratory data analysis and classification models, plus n8n workflow automation including OCR-based document processing and front-end delivery.
GitHubSend the brief today and you will have a fixed quote back within a few hours.