These 12 Deep Tech start-ups and SMEs are joining, within a joint disruptive innovation programme, an ecosystem of some thirty major groups, SMEs and research laboratories to jointly address the challenges of trusted artificial intelligence (AI) for industrial applications.

During an event organised on 16 December 2021 in the presence of its ecosystem, the industrial and academic consortium* Confiance.ai announced the 12 winning start-ups and SMEs from the call for expressions of interest launched in July 2021. AI Redefined, AIvidence, AzurIA, Cervval, Datakalab, Golaem, Jolibrain, Kereval, Mydatamodels, OKTAL-SE, Quantmetry and Simsoft-Industry will join the programme from January 2022. They will put their simulation, human-machine interaction, test and explainability technologies at the service of the development of a sovereign trusted AI engineering platform for critical systems. This programme receives €30M in funding from the Programme d’investissements d’avenir (PIA).

“Confiance.ai is a unique programme, which we are proud to participate in, as it is part of a truly integrative and open approach. With the arrival of these 12 nuggets, in less than a year of existence it has federated a rich ecosystem of more than 40 industrial partners of all sizes (start-ups, SMEs, large groups), academics and research laboratories around the common ambition of making France a leader in trusted AI”, comments Mohammed Sijelmassi, CTO of Sopra Steria and member of the Confiance.ai management committee.

“Confiance.ai is another example of the French collective at work between large groups, startups and innovative SMEs. More than ever, French Tech is an opportunity to accelerate the digital transition of the French ecosystem and meet the challenges of digital sovereignty,” says Paul-François Fournier, Executive Director, Innovation Division, BPI France.

These 12 innovative start-ups and SMEs were selected from 44 applicants by a committee composed of representatives of the programme partners on the basis of the following criteria.

  • the quality of the offer,
  • the interest of the proposal,
  • its relevance to achieving the programme’s objectives,
  • the implementation constraints,
  • the resources mobilised,
  • the funding requested

Integrated into one or more of the seven projects of the programme, they will be able to test and improve their technologies and advance them on critical use cases, while contributing to the resolution of the programme’s technological barriers. They will be supported by a financial investment of up to €1.2 million in total.

Due to the richness of the ecosystem of start-ups and SMEs specialised in artificial intelligence and the relevance of existing technologies, the Confiance.ai collective is considering launching a second Call for Expressions of Interest in the coming months.

“We are very pleased that these 12 innovative start-ups and SMEs have joined the programme, in line with the roadmap we had set ourselves. The adage “unity is strength” also applies in the world of research and innovation. At SystemX, we have proven this through our START@SystemX initiative which has, on several occasions, given start-ups the opportunity to join with major accounts and technology suppliers to build the new services of tomorrow. There is no doubt that the technologies of the start-ups selected in the framework of this AMI will contribute to accelerating the resolution of the programme’s barriers”, explains Paul Labrogère, Director General of IRT SystemX, who is leading the Confiance.ai programme.

The 12 winning start-ups and SMEs of the AMI :

AIR (AI-Redefined) : Based in Montreal, AIR develops Cogment, an open source software platform (Human in the Loop Learning). Cogment fosters continuous, real-time collaboration and learning within intelligence ecosystems that mix humans and AIs of all types.

AIvidence : Young French innovative structure dedicated to the explicability and auditability of algorithms. In confidence.ai, AIvidence will develop, using open source tools, an approach of “regional” explanations, complementary to the local and global approaches, in order to facilitate the development, between business experts and model designers, of AI that is explicable by construction.

AzurIA :  With the ambition of helping the planet and its occupants, this start-up, based in Sophia-Antipolis, optimises for each of its clients an Artificial Intelligence solution, embeddable on low-power hardware targets, to extract useful information in real time as close as possible to the sensor.

Cervval :  Cervval uses numerical simulation by merging physical modelling, statistical learning and 3D to create synthetic datasets to ensure that all cases are represented in a statistically consistent manner, including those that are rare and difficult to reproduce in reality.

Datakalab : Based in Paris, Datakalab specialises in the development of computer image analysis algorithms to measure flows in public space. In confidence.ai, it will work on improving the performance of image processing in embedded applications. The technologies of Datakalab make it possible to divide the computing power of a neural network by 6 on average.

Golaem : Golaem, based in Rennes, France, provides solutions for populating virtual environments with digital characters to feed deep learning algorithms with behaviourally accurate synthetic data. In confidence.ai, Golaem will contribute to a use case in autonomous driving.

Jolibrain : Based in Toulouse, France, Jolibrain offers the open source software product JoliGAN, which allows the transformation of data between domains with controlled conservation of semantics. This ensures the reliability of models in real context, from simulated or external data.

Kereval : Based near Rennes, this SME is a software test engineering laboratory. In particular, it develops AI verification and validation methodologies. Kereval will contribute to confiance.ai with its black-box testing and verification platform for machine learning-based systems, which allows its applicability regardless of the type of model.

MyDataModels : A start-up based in Sophia Antipolis, MyDataModels has developed its own artificial intelligence technology, ZGP (Zoetrop Genetic Programming), which is particularly effective on small data sets. The predictive models generated are extremely light, robust and operational on all types of sensors or embedded systems. The AI developed by MyDataModels is also interpretable, understandable and transparent.

OKTAL-SE : OKTAL-SE is a small and medium-sized company based in the Toulouse region of France. It is a software editor for the simulation of state-of-the-art electro-optical and radio frequency sensors. Its tools have become an international reference for the development and evaluation of critical detection, guidance and navigation applications. In the programme, OKTAL-SE will work on the production of physically realistic synthetic sensor data suitable for learning intelligent algorithms.

Quantmetry : Based in Paris, Quantmetry is a leading consulting firm specialised in Artificial Intelligence. In the programme, Quantmetry will contribute to the development of a trusted AI in France by accelerating the development of methods and tools capable of measuring the uncertainty linked to AI, in a systematic and transparent way (e.g. the MAPIE open source library).

Simsoft-Industry : Based in Labège, the start-up is developing an intelligent voice assistant called Spix. Within the framework of confiance.ai, Simsoft proposes to develop a tool-based methodology centred on SPix that will enable the establishment of trust between a technician and new AI-based systems deployed at his or her workstation and will contribute to their acceptability.

*Air Liquide, Airbus, Atos, CEA, Inria, Naval Group, Renault, Safran, IRT Saint Exupery, Sopra Steria, IRT SystemX, Thales, Valeo