Our Challenge

How can we better analyze the large amounts of data collected until today, and help the insurance, financial services and healthcare sectors invent better data intelligence ?

Our Technology

DreamQuark puts its intelligence into developing the best algorithms, to detect, very efficiently, rare and otherwise invisible phenomena in a wide variety of data types (images, texts, audio files...)

Our Solutions

DreamQuark develops a platform to facilitate access to this expertise of large datasets analysis by offering a transparent solution, with results accessible to all types of business users.

Our Challenge

Crowd image
Medical image

    Reducing costs - AI has shown very efficient results at identifying costs across the value chain. Use cases include fraud detection, churn analytics, risk pricing, product segmentation and customer qualification. By gaining a better understanding of their costs, and prioritizing accordingly, professionals will be free to take on a more strategic role.

    Targeting Emerging Risks & Product Innovation - New risks are emerging (such as cybersecurity, climate change), analyzing these trends and evaluating if there is a fitting insurance market for these risks are now natural machine learning tasks.

    Automation - AI is improving abilities in customer interaction, resolution times, and delivery speed to market of new products. This efficiency is the result of AI accelerating decision-making (automated underwriting, auto-adjudicating claims, automated financial advice).

DreamQuark envisions to help professionals benefit from the latest advances in machine-learning to create more intelligence from their datasets. Our mission is to empower the next generation of prevention, prediction and segmentation tools using data-related technologies.

Our technology

    As former particle physicists, we use our expertise of large datasets analysis and efficient algorithm design to reinvent how data analysis is performed in the insurance, financial services and healthcare sectors. We know how to deal with complex signals, and how to detect even the smallest, invisible effects. Yet, our expertise is not enough, and we believe that it is the business knowledge of our clients' experts that give the final value to the results obtained!

    Deep-learning models can be tricky to explain. We worked to develop a solution that not only comes up with results, but also explains the process through which a decision was reached. Learning which element triggered the decision will be a decisive factor in understanding your results, and focusing efforts on key parameters.

    We currently develop technologies related to deep neural-networks with sparse architectures that can unveil new patterns inside the input data. This sparcity allows for an higher accuracy and speed. This work combines our knowledge of theoretical physics and artificial intelligence. We apply these new models to both structured and unstructured data, and we transform the initial data in order to create a perfect match between the data representation and the algorithm we want to finally use. Our goal has been to combine different algorithms to benefit from the different characteristics of these algorithms and create analyses that can easily generalize. We embed these algorithms, first trained on specific datasets, into innovative applications for insurance, financial services and healthcare professionals.

Particle Physics Atlas detector collision
Deep Learning artwork

Our solutions

Everything starts from your data:

Images, time series, texts, audio recordings, Excel files... We treat datasets of all sizes, both structured and unstructured. We also automated a large part of the data pre-processing in order to obtain your results faster.

We learn from your data and your expertise:

We develop dqANN, a machine learning framework with technologies intended to mimic the behavior of the human brain. We build lighter architecture to extract pattern and features in your data faster and with an increased accuracy.

We help you play with your model:

Our platform is a White box: every decision made by the model can be explained, thanks to visualizations illustrating the variables responsible for building a specific model, a specific cluster, or even a specific profile. We help you tune parameters and explore new possibilities.

Build new strategies and products:

We help you embed custom models in your digital systems or in mobile applications. You can also use the model data generation abilities to explore new scenarios and build new strategies.

DreamUp Vision, spin-off of DreamQuark

DreamQuark created DreamUp Vision to accelerate the diagnosis of retinal diseases.

The early detection of retinal diseases could prevent blindness. DreamUp Vision's machine learning engine performs a quick detection of complicated patterns in retinal images, and is capable to detect the smallest effects, facilitating early diagnosis of different retina disorders. Deep-learning allows this detection with the performances of professional ophthalmologists in milliseconds. We have implemented this engine both in a mobile application and an online web-application.

To learn more about our project, please visit DreamUp Vision

retina image

Our references

They use our solutions:

Malakoff Mederic
LLOOPP - Natural intelligence solutions
AXA Global direct

They believe in us:

Telecom ParisTech
La FrenchTech
CCI Cote d'Opale
Paris Region Lab
Institut Open Innovation

Our Unique Team


Collision Conference Las Vegas 2015
Web Summit Dublin 2014 and 2015
Deep Learning Summit Re-Work Conference San Francisco 2015

Happy Hours de l'Innovation

Sept 17, 2014

SALONCREER2014, Lille ''La santé et l'e-santé''

Salon big data paris 2015
Hacking Health Camp Strasbourg 2015
Hackathon e-sante Paris 2015
Deep Learning Summit London 2015
NVidia GPU summit 2015
IT Innovation Forum
Collision Conference Las Vegas 2015
Web Summit Dublin 2014 and 2015
Deep Learning Summit Re-Work Conference San Francisco 2015
Salon big data paris 2015
Hacking Health Camp Strasbourg 2015
Hackathon e-sante Paris 2015
Deep Learning Summit London 2015
NVidia GPU summit 2015
IT Innovation Forum


Winner of 2 prizes at the 2015 IT Innovation Forum:

Lauréat Big Data - IT Innovation Forum

Big Data

Lauréat Meilleure StartUp - IT Innovation Forum

Best Start-Up

Lauréat Meilleure StartUp - IT Innovation Forum

Most Innovative Start-Up of the Year

Acceleration programs

Digital disruptors

Allianz accelerator



  • A CUSTOMER: If you are facing a challenge with the analysis of your data, and want to have a dedicated team eager to develop cutting-edge technologies in machine-learning and data analysis, and apply these technologies intelligently on your datasets, then you have found the right company ! Just send us a line to explain your problem, we will soon come back to you to help you tackle your challenge.
  • AN INVESTOR: If you want to invest in our company, we are looking for investors who share our values. Any investor eager to participate to our endeavour will be warmly welcome.
  • A TEAM MEMBER: We are looking for bright, enthusiastic and highly-motivated individuals with a passion for complex challenges. If you feel you have the grit to join our team, then contact us.