Case Studies: anonymized, under NDA
Real work, told honestly: we show the shape of each build.
How to read these
Our client work is under NDA by design. That is not a way to hide thin proof; it is how we protect the people who trust us with their systems. So on this page we are concrete about the things that matter to you (the sector, the problem, and the approach) and we withhold only the identifying details: the client, the product, the country beyond a region word, and the logo.
We describe capabilities and architecture, never a legal outcome. We do not claim a certification, clearance, or eligibility we have not earned. Each engagement runs through our gated delivery process: requirements, design, build under test, review, and a controlled release, with a human approving every gate.
On any regulated case we add a Note: line that states plainly what the software does and does not do: it supports your obligations and keeps a person in the decision; it never certifies, clears, or replaces you.
- Sector:
- European airport operator (under NDA)
- Problem
- Passengers and ground staff were working from flight information that lagged reality, and questions arrived in more languages than any help desk could staff around the clock.
- Approach
- We built a live flight-data platform that continuously collects operational data and republishes it to the website and passenger apps, alongside the things travellers actually ask about: transport, parking, weather, and waiting time at the checks. On top of it runs a multilingual voice-and-text assistant that answers from the real, live data rather than a canned script. Delivered on our gated process, so each release was reviewed and traceable.
- Sector:
- European tourism operator (under NDA)
- Problem
- The operator was giving away margin to third-party booking platforms and losing customer requests in an inbox nobody could keep up with.
- Approach
- We built a direct-booking website and a partner portal, and put an automatic triage step in front of incoming customer requests so each one is categorised and routed the moment it arrives. Availability and bookings stay in sync with the tools the staff already use, so the direct channel never sells what is not there.
- Sector:
- Public-sector body (under NDA)
- Problem
- A compliance process that leaned entirely on documents (checking staff declarations, assembling case files) was run by hand inside a strict regulatory frame, and it did not scale.
- Approach
- We automated the document workflow on the body's own infrastructure: nothing leaves the organisation. The system reads and assembles the paperwork, every result is linked back to the source that produced it, and a person can override any step. Data residency here is a genuine architectural answer, not a promise that data "stays in the region".
- Note:
- The system supports your compliance obligations and keeps a human override at every decision; it does not certify or guarantee compliance. That determination stays with you and your regulator.
- Sector:
- European defence programme (under NDA)
- Problem
- Partners on a collaborative defence programme needed to publish and test reusable software components without exposing them to one another and, separately, to spot new construction in aerial imagery and tell what had actually changed.
- Approach
- We built a partner platform that publishes and tests components inside isolated environments, so each partner sees only what it is cleared to see. Alongside it, an applied-vision pipeline detects buildings in aerial imagery and compares every new detection against what is already on record, surfacing only genuine changes for an analyst.
- Note:
- We are an EU-established partner; programme eligibility depends on establishment and control tests and on the applicable export-control rules. We describe our role factually and never self-assert a clearance.
- Sector:
- Weather and marine forecasting service (under NDA)
- Problem
- A forecasting service needed atmospheric and sea-state forecasts delivered in the exact formats its geographic-information tools already consume, without a manual conversion step in the middle of a time-critical cycle.
- Approach
- We built a pipeline that runs, on a fixed cycle, an atmospheric model and then a wave model for height, period, and direction, and exports the results straight into the standard GIS formats the service already uses. Repeatable measurement is built in, so each run is comparable to the last.
- Sector:
- Agriculture and rural-technology provider (under NDA)
- Problem
- Growers needed a plant-disease diagnosis while standing in the field, often with no network at all, and the provider needed early warning of outbreaks across a region.
- Approach
- We built an app that photographs an affected plant and returns a diagnosis in seconds from a model running on the device itself, so it works fully offline. When the model is not confident, the case is routed to a human agronomist rather than guessed, and results across a region roll up into outbreak alerts.
- Sector:
- HR and candidate-screening team (under NDA)
- Problem
- A hiring team needed to separate two questions that are easy to conflate: whether a candidate can do the job, and whether what they claimed about themselves is actually true.
- Approach
- We built a screening assistant that keeps those questions apart, matching fit to the role on one side, and checking the verifiable trail on the other (identity, genuine credentials, a consistent history), and attaches the supporting evidence to every signal. The assistant explains each finding; a person makes the decision.
- Note:
- The assistant assists human review and does not make hiring decisions. Candidate-screening AI is a high-risk use; human oversight is built in and the hiring decision stays with your team.
- Sector:
- Retail-investor financial service (under NDA)
- Problem
- A service for everyday investors wanted to answer common market and portfolio questions grounded strictly in its own approved material, not in whatever a general model happened to have absorbed.
- Approach
- We built a digital advisor that answers in real time from the organisation's own document set, so every answer traces back to an approved source rather than a black box, with portfolio tracking and selected alerts. It is available on both the web and a mobile app.
- Note:
- The advisor provides general information, not investment advice or a personal recommendation; content is kept genuinely generic and the decision remains the investor's own.
