Molecular and Atomic Spectroscopy is a well-established set of technologies that use the electromagnetic spectrum to generate a unique fingerprint of the molecular structure, from the atomic scale to that of molecules, proteins, mixtures, humans, cities, forests, continents and oceans. Though well-established, the Molecular and Atomic Spectroscopy market sees rapid development related to both technology (miniaturization of components using e.g. MEMS technology and massive increase of embedded processing power) and to expanding needs in its use areas (pharmacy moving from classical chemistry to synthetic biology, the need for personalized medicine, and omnipresent security and environmental surveillance).

The overall objective of Spectralligence is to accelerate innovation and growth of the Molecular and Atomic Spectroscopy market segments by augmenting innovations in micro-electronics and component miniaturization by cross-domain validated neural networks for spectral analysis. These industrial-grade embedded AI technologies are crucial enablers for one-click spectral characterization in materials and life sciences. Resulting implementations will advance adoption of currently highly expert-dependent applications by Cost Optimizers and Standardizers, esp. by cost-effective and easy-to-use products.

Spectralligence will develop and leverage Neural Networks to augment and replace expert users in data workup for the different technologies in Molecular and Atomic Spectroscopy. We bring together 2 Industrial Partners (Philips and Spectral Engines), 2 SMEs (Olfactomics and Dynaxion), and 2 academic partners (TU/e and MUMC) with expertise on deep learning and the necessary deep physics and application knowledge base necessary to develop AI solutions which are robust against domain shift whilst delivering superior performance in each specific domain. Our focus is on innovative sensing applications enabled by dedicated miniaturized components. Cooperation in Spectralligence will advance the technological level of each of the partners in the field of AI and strengthen the industrial partners’ market position, enabling them to develop and offer novel, easy-to-use and affordable spectral analysis solutions tailored to the needs of non-expert users.

Spectralligence will apply AI and Deep Learning to a broad range of spectroscopic applications and ever-increasing amounts of data generated with novel miniaturized sensors, to address two aspects: (i) robust data (pre-)processing and spectral denoising, and (ii) feature extraction from the resulting complex spectroscopic data sets. The results must be trustworthy with a high level of confidence in the domains of pharmaceutics, toxicity, medical diagnosis and (explosives) safety. Embedded AI solutions will be rigorously developed to ensure compliance with standard industrial and regulatory processes, including Verification, Validation, Quality Assurance and Quality Control and Accreditation. The technologies and products developed in Spectralligence aim for end-customer market segments with double digit CAGR. These high growth rates are expected to be sustained irrespective of the COVID-19 economic burden: we focus on segments related to healthcare and medicine, security, and environmental care. As a direct impact of the AI technologies developed in Spectralligence, the commercial partners anticipate to generate a total annual revenue increase of 15-25 M€ within 3 years after the end of the project.