Data agents that already run inside real companies.

Three recent builds, from customer-facing AI to internal search. Details anonymised at our clients' request.

Customer-facing RAG 01 / 03

Shoppers ask.
The catalog answers.

Product Q&A for a large clothing webshop

Customers were leaving product pages to hunt for sizing, fabric and care details buried in descriptions, reviews and supplier sheets. We built a retrieval-augmented generation model over the full catalog, so shoppers ask direct questions about any product and get a grounded answer with the source shown.

  • Catalog and review indexing with nightly sync
  • Grounded answers with citations, no made-up specs
  • Evaluation suite run on every release
Email agent 02 / 03

50% of inbound email answered fully automatically.

Autonomous email replies for a support team

Half the inbox was routine: order status, returns, invoice copies, address changes. Our agent reads each email, pulls the facts from the order and CRM systems, and replies on its own when it is confident. Everything else is drafted and routed to a person with the context attached.

  • Connected to order, CRM and ticketing systems
  • Confidence gate decides between auto-reply and human handoff
  • Full audit trail of every sent reply
Enterprise search 03 / 03

The right document on page one, not page five.

Relevance overhaul for an internal document search

Employees had stopped trusting the company search: keyword matching surfaced outdated and duplicate files ahead of the one they needed. We rebuilt ranking around semantic retrieval, recency and document ownership, and measured relevance with real queries from the teams using it.

  • Hybrid keyword and semantic retrieval
  • Ranking tuned on real user queries and click data
  • Respects existing document permissions

B2B data consulting

We build the data agents your company runs on.

Datalumina is a consultancy for businesses that want to put their data to work. We design and build data-driven products for your teams, mostly for internal use: from customized RAG models that answer questions over your own documents, to full end-to-end webshops when the product faces your customers.

  • ✦ RAG assistants
  • ✦ AI agents
  • ✦ Search & ranking
  • ✦ Data platforms
  • ✦ Webshops

Services

What we build

Products for your teams first, for your customers when needed. One team, from the first data model to the last pixel.

01

Internal knowledge assistants

Customized RAG models over your documents, tickets, contracts and wikis. Grounded answers with source citations, access control and evaluation built in.

02

Data platforms and dashboards

Pipelines, warehouses and reporting that bring scattered company data together, so management and operations work from the same numbers.

03

Internal tools and automation

Search, document processing and agents woven into the systems your teams already use, with the guardrails a production system needs.

04

End-to-end webshops

Catalog, checkout, payments and fulfilment, connected to your internal data so pricing, stock and recommendations stay in sync.

Approach

How we work

Consulting that ends in working software. Small senior team, short feedback loops, and code your company owns.

  1. 01

    Discover

    We sit down with your teams, map the data you have and the problem it should solve, then agree on a scope that ships.

  2. 02

    Build

    Weekly releases your colleagues can click through, with measurable quality for every model and feature.

  3. 03

    Run

    We deploy into your environment, hand over with documentation and training, or stay on as your data partner.

Contact

Let's talk about your data.

Tell us about the challenge your company is facing, or the data you are sitting on. We will tell you honestly what we can build with it.

We reply within two business days.