The open questions behind
structured intelligence.
Djeed Lab is our applied-research arm — and our agenda in the open. The hard, real-world questions we work on to make AI trustworthy at scale: how scattered records become one shared model, how the same entity resolves across systems, how every number stays traceable, how to read place and foresee what comes next. Applied, not academic — each question has a home in our solutions and shows up in real use cases.
A distributed team · Lausanne — Lisbon — Cairo
The thesis
The hard part isn't the model. It's the structure beneath it — how a domain is modelled, how scattered records resolve into one truth, how a number stays traceable, how a forecast earns its confidence.
The Lab is where we take those layers apart and work out methods rigorous enough to build on. This is our agenda, in the open — and we are actively looking for the people who want to work on it.
§ Research agenda
Five questions we keep coming back to.
Each is a genuinely open, applied problem — and each has a home in our solutions and a place where it is already being used. The topics below are where we work; the published findings will follow.
How do you turn an institution's records — exactly as they are — into one shared, evolving model, without asking anyone to change how they work?
The hard part — Drifting taxonomies, heterogeneous formats, and a schema that has to evolve mid-project — while provenance survives every change.
Connects to
When the same person, place, or asset appears across systems that share no key, how do you resolve it into one entity you can trust — and prove every link?
The hard part — No shared key, truth that arrives asynchronously, and merges that must stay auditable at scale — sometimes without moving the data at all.
How does a single number on a decision-maker's screen stay traceable to its source — through every roll-up, forecast, and AI-written summary?
The hard part — Provenance breaks exactly where it matters most — at aggregation and at generation. Keeping it intact separates a finding from an assertion.
How do you read where and when something is happening — fairly — when the signal itself is unevenly collected, and anticipate where the next one forms?
The hard part — Maps and forecasts inherit the bias of how their inputs were gathered. The work is correcting for that honestly — and still seeing what's coming early enough to act.
How can a region's natural and economic signals be structured into an asset it can plan, invest, and build around?
The hard part — Fusing many incompatible public and sensor sources into one investment-grade, provenance-clean, license-aware layer — from a single site to a national register.
§ How it all connects
Not five subjects — one system.
These aren't five separate threads. They're stages of one arc — understand a domain, connect what's scattered, keep it trustworthy, anticipate what's next — and a finding in one moves the others. A question worked out in the Lab becomes a method in the foundation, ships as one of the five solutions, and shows up in a real use case.
§ Where we work
Distributed by design.
The Lab operates across three cities and the time zones between them — European research culture with Mediterranean and North-African reach, from the Alps to the Atlantic to the Nile.
Lausanne
Switzerland
Swiss base — rigour and residency
Lisbon
Portugal
The Atlantic edge of Europe
Cairo
Egypt
Regional reach across the Mediterranean and beyond
Working on one of these? Come find us.
We're looking for researchers and practitioners — in machine learning, data engineering, geospatial science, urban and tourism economics — who think about these problems. If one of these questions is yours, or you're holding a dataset or a problem that looks like one, write to the Lab. The published catalogue will follow; the conversations start now.