EMAIL: contact@disaitek.ai
Who we are and what we do
Disaitek's purpose is to fight against environmental and public health damages. Our goal is to get rid of pollution, make companies comply with environmental regulation and shed light on unpunished polluting activities.
Founded in 2017, Disaitek is a French company based in Paris which has invested early in R&D to develop strong skills and a best of its kind knowledge in the integration of optical, thermal and SAR satellite data into proprietary AI models to raise in semi real time alarms about pollution events at risk (illegal landfills, GHG leaks, illegal lodging, pollutants, etc.)
Our cutting-edge algorithms automate the detection of pollution sources at scale, allowing us to notify end users about risks like fire outbreaks, air, water and soil pollution and public health risks.
The new space sector with artificial intelligence will disrupt the environment risk market and environmental compliance market. With its abilities to deliver at scale, targeted markets for Disaitek are worldwide and benefit large corporation with interests all over the world.
Disaitek specializes in the detection, quantification and monitoring of sources of pollution by leveraging earth observation technologies, advanced artificial intelligence and open source intelligence.
Disaitek is labelled as French Young Innovative Enterprise.
Disaitek is part of the incubation program delivered by ESA and Incuballiance (Paris Saclay).
Disaitek has also been accepted as part of the BLAST acceleration program (consortium composed of Polytechnique, SATT Paris Saclay, ONERA and Starburst, with CNES as partner).
Skill set
Computer Vision
In Image Recognition, our core skills have been strengthened with one outstanding experience we have conducted in 2020 with the University of Geneva in order to develop new methodologies to detect exoplanets. Deep learning applied to exoplanets has been here for a few years but only for uniform orbital transit. Using data from the Kepler & TESS space telescope, our challenge was to detect exoplanets with non periodic orbits. We have implemented state of the art image analysis technology in a context of low signal to noise data to be able to achieve this task.
Since then we have developed new methodologies using self and semi supervised training in order to reduce the amount of labels required for semantic segmentation tasks.
Earth Observation
Disaitek uses state of the art processing methodologies to combine data coming from different sensors and satellites. We leverage optical, thermal and SAR data to train AI models in various methodologies (SSL, semi supervised and supervised) to capture insights on what we observe in the air and on the ground.
Natural Language Processing
Disaitek area of expertise is the use of deep learning methods for NLP tasks. More specifically, we use neural networks that have been trained in an unsupervised way on a huge corpus of documents and transfer their knowledge to other problems. In order to do this, we use state of the art semi-supervised learning algorithms and create proxy tasks to get the most value possible out of our data.
Disaitek has been contributing to Pytorch implementations of GPT and BERT, large pre-trained models that capture deep information about syntax and grammar of languages.
Solutions
Main clients
Partners
European Space Agency
Incubation Program
Disaitek has been selected to be part of the ESA incubation program to help our roadmap progress. It is a recognition of usefulness, our innovation and our ability to execute our vision. Along with Incuballiance (Paris Saclay), they provide highly valuable network ressources (Funders, technical expertise with ESA & CNES and Business development support with ESA)
BLAST / STARBURST
Acceleration Program
Disaitek is part of the BLAST acceleration program which is a consortium composed of Polytechnique, SATT Paris Saclay, ONERA and Starburst. The CNES is a partner of the program.
We benefit from high expertise on SAR and GHG sensors, IP assessment and registration. We have access to international experts, mentors and funds.
CORE TEAM
Elise Colin
Scientific Advisor
Research Director ONERA
Anthony GRAVELINE
CEO / Co-founder
Civil Engineer
Gregory Chatel
Lead R&D / Co-founder
PhD Computer Science
Flavien Vansyngel
Lead Science
PhD Astrophysics
Rodrigo Andres Rivera Martinez
PhD Climate Physics
Thibaud Richard
PhD Astrophysics
ENS
We’re Hiring!
PhD Computer Sciences, Mathematics or Physics:
Doing research and development for a safer world
Data Scientists:
We want to leverage continuous learning from users to optimize our deep learning stack and to personalize the way the systems act regarding the user habits
Developpers:
We want to build great experiences for our users and let them feel comfortable with our solutions. Great user experience for great people (technology .js)
Let’s Meet UP!
Disaitek has developed strong expertise in driving R&D works….We propose to share it!
Meet UP has become a nice way to gather people and expert pleased to explain and share insights on lots of discipline.
Traditionally this meeting takes place in a remote place, gathering people from different horizons (employees, independent, founders, VC, academics…). Usually these people act outside of their work hour and independently from their company or others.
We think company must leverage this kind of activity and bring it inside your company
Why: It creates condition to innovation: Giving new insights to employees, they can imagine applicable solution on their work and in the process they are involved in
-
It creates conditions to remove silos inside the company. The time of the event, people can meet people from other services, create new internal links, and find new way to collaborate towards customer,
-
It is also an opportunity to identify specific talents, and imagine new opportunities for internal mobility,
-
Finally, it can help build an innovative and positive image of the company to the employees, which in the current period of talent shortage is valuable.
Disaitek offers to bring valuable meet up inside your company. Depending on the level of the attendants you program to invite, several agendas are possible.
If you are interested in bringing information to your employees on breakthrough technology, if you want to experiment new engagement forms with them, contact us. You just have to provide snacks and refreshment and we handle the presentation for free!
Meetups can be done either in English or French.
Which Meetup is better for you
Introduction to fundamentals of Machine Learning
Ok, what’s really behind the scene?
Who ?
All kind of employees curious about this phenomenon and who try to find new tools to increase productivity in their work.
Agenda
-
Give better intuition about what covers Machine learning
-
Categories of machine learning
-
How to train a machine learning model
-
When it comes to evaluate the precision of your model
-
Relevant frameworks of machine learning and AutoML
-
Bias in machine learning
-
Machine learning for NLP
-
Recognize context in your day to day job where machine learning can apply
Takeaway
Help to come up with relevant use cases suitable for your company
Unlock the potential of your Data
Hard to have enough labelled data?
Who ?
Data scientist, people involved in machine learning development and concerned by the low quantity of labeled data.
Agenda
-
How to train deep learning models on language modeling helps capturing syntax and semantic of languages
-
How to use to transfer this knowledge to other tasks
-
Multi-task learning: principles
-
How multi-task learning help models generalize better
-
Principles of semi supervised learning
-
Example of Virtual Adversarial Training to reduce the need of labeled data
Takeaway
New ideas to tackle existing problems
Machine Learning privacy and security
How really secure and ensure your models respect customer privacy?
Who ?
All people involved in machine learning development
Agenda
-
Key dates for this discipline
-
Principles of security and privacy of machine learning
-
State of the art in machine learning security and privacy
-
Backdooring
-
Poisonning
-
Model stealing
-
Data stealing
-
Principles of prevention. security and privacy by design
-
Watermarking
-
Differential privacy
Takeaway
First evaluation of the risks of currently deployed models or systems