Skip to content
Part of the Equinox Group · Malta, EU · Mauritius, Africa
equinoxgroup.eu

Insights

Insights and publications

Guides on AI regulation, technical notes from our EU research and practical perspectives on automation and language models, written by the people who build the systems.

Latest

Guides and technical notes

Data and interfaces · 6 min

Mauritius Water Twin: what public data say about the island’s water

An independent map of Mauritius, hexagon by hexagon. Every value says what kind of fact it is and names its source and date, and what is missing is shown, with who holds it.

The twin records the gap; it does not assign blame.

Read

Research and forecasting · 5 min

Where food crises are likely next: FP-EWS forecasts on a 3D globe

Forward forecasts of food-crisis risk for 507 provinces in Africa, Afghanistan, Guatemala and Haiti, each frozen before the maps that will score it exist. What they show, and how they were tested.

Results are reported whether they pass or fail.

Read

Data and interfaces · 7 min

An ocean of records: the Copernicus Ocean State Report on an interactive 3D globe

With a possibly record El Niño building, the tenth Copernicus Ocean State Report as an interactive 3D globe, drawn live from Copernicus Marine data. What it takes, and what a map can and cannot claim.

Where the map is a model rather than an observation, it says so.

Read

Data and interfaces · 7 min

Below 10 metres: EU crop diversity on an interactive 3D globe

Open EU data on what grows where, rebuilt as an interactive 3D globe and taken from 10 m satellite pixels down to the fields farmers declare. What it took, and where the comparison stops holding.

Regional averages can hide exactly the fine-grained structure that diversity is about.

Read

Strategy · 6 min

From AI use to AI value: where small businesses should start

Most small firms already use AI. Few have built it into how the business runs. Why that gap exists, and four moves that close it: start from the bottleneck, redesign the workflow, set guardrails people can follow, and measure what changed.

Small businesses have little room for technology that looks impressive and changes nothing.

Read

AI Act · 8 min

The AI Act after the Omnibus: a readiness guide

The high-risk deadlines moved. Most of the Act did not. What changed with Regulation (EU) 2026/1744, which dates now matter, and five steps to get ready.

It is tempting to read this as a pause. It is not.

Read

AI Act · 6 min

AI literacy after the Omnibus: from “sufficient” to supporting measures

Article 4 was rewritten in July 2026. The duty is now one of effort rather than result. What that means in practice, and what a defensible literacy programme looks like.

It moved from a result to be achieved to an effort to be made.

Read

Technical note · 7 min

Seeing what plastics hide: multimodal defect detection

Why 2D vision struggles on moulded and extruded plastic parts, how 3D profiling and thermal sensing close the gap, and what it takes to fuse them in real time on a robot.

Inspection stops being a gate and becomes the first half of a repair loop.

Read

Technical note · 7 min

Virtual sensing at the edge: physics-informed estimation for mould changeovers

Some of the most important quantities on a production line cannot be measured cheaply. How a physics-informed virtual sensor can estimate them from commodity vibration sensors, on a microcontroller, with known error bounds.

The operator gets an answer with a stated error bound rather than an unexplained score.

Read

Automation · 6 min

From RPA bots to AI agents: when to upgrade

Rules-based bots are cheap, fast and auditable, until the work stops being predictable. The signals that a process has outgrown RPA, the cases where it has not, and a hybrid pattern that gets the best of both.

Each layer does what it is best at, and the overall process stays auditable.

Read

LLMs · 6 min

Private LLMs: on-premise or API?

Commercial model APIs and self-hosted open-weight models each have a place. The factors that should drive the choice, and why many organisations end up with both.

The pragmatic answer is often a hybrid.

Read

Publications

Research by our team

Peer-reviewed and preprint work by members of the Algorithmic BrAIn and Equinox Group team.

AI-Guided Self-Driving Laboratories for Advanced Materials Discovery

Bernard Mallia, Rossana Caputo · Computing & AI Connect · Journal article · 31 August 2026

An Energy-Efficient, Privacy-Preserving Distributed Ledger Framework for Agricultural Equipment Traceability Using Passive NFC

Rossana Caputo, Bernard Mallia, Keerthi Masanasetty, Prasad M · Preprint, 2026 · DOI 10.2139/ssrn.6433100

Perspective Chapter: The Faint Whispers of the Heart – Next-Generation Sentiment Analysis and “Empathy Algorithms”

Bernard Mallia · In J. Li (ed.), Emotions in Code – The AI Frontier of Sentiment Analysis, IntechOpen · Book chapter, 2025 · DOI 10.5772/intechopen.1010889

Designing and Implementing Socially Beneficial Recommender Systems: An Interdisciplinary Approach

Bernard Mallia · NORMalize 2023, First Workshop on the Normative Design and Evaluation of Recommender Systems, co-located with ACM RecSys 2023, Singapore · CEUR Workshop Proceedings, Vol. 3639

Have a question we should write about?

Tell us what your organisation is working through. The best topics come from real problems.