Education and research

What is at play in education and research

Knowledge institutions manage hundreds of thousands of sources: publications, reports, teaching material, datasets. Keyword search finds only part of that. At the same time openness and ownership weigh heavily: publicly funded knowledge should not be locked into a closed platform, and research data does not simply leave the building. That calls for search technology you control yourself.

Working in a strict sector calls for demonstrable information security. Gradient is ISO 27001 certified.

Where it pinches

Keyword search falls short
Anyone who does not know the exact term will not find the document. Semantic search also finds what is worded differently.
Ownership of data and technology
Knowledge institutions do not want to depend on a single vendor or a single cloud. Open source and self-hosting are often a requirement.
Scale
A knowledge library of hundreds of thousands of sources has to answer within seconds, even as the collection keeps growing.

What we do

All services

01
Semantic search at scale
We make large knowledge libraries semantically searchable, with a reference to the right passage. That way you also find what is worded differently from your search term, within seconds.
02
Open source, under your own control
The solution runs on open-source technology in an environment the institution decides on. That way the organisation stays the owner of data and technology, without dependence on a single vendor.
03
Measuring whether it improves
We establish in advance what a good search result is and measure against that. That way an improvement is a number and not an impression.

Our work in education and research

All cases

GKN · WUR and Groen Kennisnet

Making research material searchable.

Semantic search across the Groen Kennisnet knowledge library, fully sovereign and on open source.

MIT Feasibility · RVO

Improving compact language models with reinforcement learning.

A feasibility study with MIT funding via RVO into the targeted improvement of compact, open-source language models with reinforcement learning. We tested both the technology and the market demand.

Trusted by