"Machines aren't the thing,
they're the thing that gets us to the thing"

Joe McMillan, "Halt and Catch Fire"

A small core crew of AI and Web veterans with connections. We approach our projects like a Miles Davis band, valuing competence, a good sense of humor, originality and collaboration.

Leave the AI hype on the doorstep, we make real applications that work. We run a reality-check on the toughest challenges, and if we decide to take on it, it means we are certain to succeed.

We've reached 100 projects in 2024: many large corps, scale-ups and start-ups work with us. Our methodology has been battle-tested.

We apply a 3 short steps methodology that proved successful in our ML projects

- Understand the details and intricacies of the industrial task to be automated, with our customers and their employees. Every detail and data matter.

- Re-formulate the task as ML to best fit all constraints (costs, low data, inference time, ...). New AI techniques do unlock higher automation once the problem is framed properly. This is where we shine by bridging the gap between academia and production.

- Execute and deliver, break boundaries when needed with focus to building usable solutions, not proofs.

We provide compute, software and senior AI experts, and migrate everything on customer' premises once ready.

We build high end production software that embeds the latest AI advances with rigor and creativity. Our software runs on the premises of very large corporations and startups alike.

We integrate and build AI for real-world applications. We're experts in Machine Learning, Deep Learning, Reinforcement Learning, Stochastic Optimization, Search Engines, P2P systems and more.

We make beautiful applications out of maths: Data Visualization and Exploration through Web technologies, Big data exploration with VR, embedded hardware and installations.

The industries that trust us are eclectic because AI is pervasive. They range from large cybersecurity players to top art institutions, AI hardware startups and manufacturers, aerospace leaders and machine learning startups around the globe.

Our experience is clear: bullet-proof software is Open Source. We like creating, contributing, learning and sharing. Most of our work is publicly available and Open Source.

Deep Learning as a Service: we are the makers of the Open Source Deep Learning platform and server DeepDetect, generative AI toolset JoliGEN for images, and next generation scheduling under uncertainty Wheatley.

Aerospace and Transport Industries: Top corporations use our custom machine-learning based systems in planes, trains and cars to increase both safety and autonomy.

Art: Tate Recognition exhibit runs our code.

Fashion and Luxury: World-class leaders run custom-trained Augmented Reality and Virtual-Try-On models for Web and in-boutique experiences.

Defense and Space: We collaborate on challenging optimizations for Earth and Space applications. Space is the place.

Manufacturing and Supply Chain: Aerospace suppliers and leaders use our reinforcement learning based approach to scheduling, that is the first to readily handle uncertainty.

Cyber-security: Some mega-corps run our deep learning service for malware detection. Some others detect anomalies in logs thanks to our custom application.

Content & Media: Some large corporations and startups run our deep learning server for content filtering and moderation.

Images & Search: Some e-shops and aerospace agencies search and recommend visual elements running our pipeline.

High tech & hardware: Some deep learning startups run their cloud services on our code.

We're based in Toulouse, France, and we like it like that.