Research

Questions worth testing.

VERITAS-CA proposes research into trustworthy cognitive AI. NODERIQ-RCN explores collaborative mission intelligence; QANTIS, qmesh and QMANN investigate distinct quantum questions.

Start with the question, then inspect the method and public record. Each research route states what its results can support and where evaluation still stops.

Conceptual sculpture: three separate viewpoints around a shared task, with a gap in a blue connecting line — Conceptual study of different viewpoints around a shared task.

Horizon Europe / Submitted proposal

VERITAS-CA

How can AI agents coordinate when information is incomplete?

Reasoning, planning and reviewable decisions under limited communication. Submitted on 28 October 2026 to the EIC Pathfinder DeepRAP call; submission is separate from any evaluation or funding outcome.

Explore VERITAS-CA

Collaborative R&D

NODERIQ-RCN

How does a bounded work order survive incomplete context or a broken connection?

NODERIQ-RCN is the collaborative R&D initiative alongside the NODERIQ product. Consortium development continues for broader applications, platform-specific integration and physical validation, targeting fragmented context and difficult multi-device handoffs.

The classical and edge path must remain useful without continuous cloud access or quantum hardware. Physical safety and consequential authority stay with the device and accountable operators.

Read the record · Current NODERIQ product brief and the separate RCN research direction; platform validation and acceptance retain their own evidence requirements.

Explore NODERIQ-RCN NODERIQ product

Stationary inspection rover and separate manipulator in a dark engineering studio, with a sensor module and closed notebook in front — Conceptual distributed robotics setting

Public quantum research

Read the method as well as the result.

Each programme has a different unit of evidence. A promising example begins a comparison; it does not settle performance or availability across other settings.

Chip package and measurement leads on a dark bench with a cryostat behind glass — Conceptual quantum measurement setting

Research question

QANTIS

Can quantum hardware estimate rare evidence for a classical planner?

A classical planner supplies a prior and observation model. The bounded quantum step estimates an evidence term, then returns an ordinary posterior for a classical decision check.

Explore the research

Method: Classical input → evidence estimate → classical return
Public record: Two public arXiv preprints and an MIT-licensed Community Edition.
Boundary: The papers do not establish a total-runtime advantage or a deployed autonomy system.

Separate chip, optical assembly and fiber interface samples in a precision tray — Conceptual compilation and modality study

Research question

qmesh

How can a quantum program keep its compilation route reviewable?

A typed program describes the work before a compatible route is chosen. Compilation, scheduling and signed manifests make the route and its resulting record inspectable.

Explore the research

Method: Typed program → compatible route → signed manifest
Public record: Apache-2.0 source and a documented compiler research path.
Boundary: qmesh can compose with other research; that does not mean every published run used it.

Protected chip fixture, unbranded compute module and blank workstation in a dark laboratory — Conceptual hybrid-compute research setting

Research question

QMANN

What do quantum-memory experiments show in each execution mode?

Theoretical work frames a model, simulation tests a defined setup, and a cost estimate precedes any paid hardware run. Results must be read within the mode that produced them.

Explore the research

Method: Theoretical → simulation → cost-gated hardware
Public record: Apache-2.0 library with three documented research modes.
Boundary: A hardware option is an experiment path, not proof of a measured benefit.

Start with quantum basics