BIRALAB

Research

BIRALAB builds one technology core — Raman spectroscopy combined with explainable artificial intelligence — and deploys it across three application layers.

Core technology

A single processing chain, from physical sample to a conclusion the end user can interrogate.

BIRALAB processing pipelineA four-stage chain: the sample is measured with a Raman spectrometer, the resulting spectrum is processed by a deep learning model running at the edge, and an explainable AI layer produces a conclusion together with the wavenumber regions that support it.
SampleNo destruction, no reagents
Raman spectroscopyMolecular fingerprint across wavenumbers
Edge AIOn-device inference, no data leaves the site
XAIConclusions with spectral evidence
Raman shift (cm⁻¹)
A four-stage chain: the sample is measured with a Raman spectrometer, the resulting spectrum is processed by a deep learning model running at the edge, and an explainable AI layer produces a conclusion together with the wavenumber regions that support it.

Four design principles

  • 1.Non-destructive

    Raman measurement relies on scattered light: the sample is neither consumed nor altered. The same specimen can be measured repeatedly or passed on to a reference method.

  • 2.Edge processing

    Models are compressed to run on the measuring device itself. Results are available at the point of sampling, independent of connectivity, and no data leaves the facility.

  • 3.Explainable

    Every conclusion is reported together with the wavenumber regions that drove it. This is mandatory in clinical and regulatory settings, where a result must withstand expert review.

  • 4.Low cost

    We prioritise commodity instrument configurations and open-source software so that provincial laboratories and small enterprises can deploy the technology.

Four research areas

One shared core of Raman spectroscopy and artificial intelligence, three distinct application layers — as set out in our Strong Research Group recognition decision — and an underpinning strand in photonics and instrumentation that builds the measurement tools themselves.

  • Healthcare

    Healthcare and biomedical diagnostics

    Non-invasive screening for skin cancer and metabolic disease, and clinical decision support.

    Learn more →
  • Pharmaceuticals

    Pharmaceutical quality control

    Active ingredient quantification, content uniformity testing, and rapid detection of counterfeit medicines and cosmetics containing banned substances.

    Learn more →
  • Agriculture

    Smart agriculture and food safety

    Pesticide residue analysis, quality control of agricultural exports, and mycotoxin detection.

    Learn more →
  • Photonics

    Photonics and instrumentation

    Integrated photonic components, Raman probes and optical measurement instruments — the engineering layer beneath all three application areas.

    Learn more →

Measurement capability and instruments

Measurement capability and instruments

The group operates Raman spectroscopy systems using excitation near 785 nm for biological samples and 532 nm for inorganic samples, paired with edge computing hardware for on-device inference. Full instrument details and acquisition conditions are published alongside every dataset in the Data Hub to keep results reproducible.

The BIRALAB preprocessing standard

Raman analysis results depend heavily on preprocessing. The group publishes an open preprocessing standard — baseline correction, intensity normalisation, cosmic ray removal, smoothing — together with source code, so that other groups can compare results on a common basis.

The BIRALAB preprocessing standard

Next step

Bring us your problem

For companies with a concrete testing problem.

Bring us your problem

Propose a research collaboration

For research groups and hospitals seeking joint research.

Propose a research collaboration