THE COMPANY BEHIND NEURYTIKA
Aerospace engineering, embedded AI and explainable prediction—connected by the same ambition: build technology that works in the real world.
Celeste EcoFlyers develops electric aircraft with pneumatic wings and the intelligence to support useful missions. Neurytika is an important branch of that work: making structured data actionable through predictions and evidence people can inspect.
TWO CONNECTED ENGINEERING BRANCHES
Aerospace puts practical constraints at the center of engineering: mass, energy, operating conditions and the information available when a decision is needed.
Our aircraft and AI activities meet around those constraints. Pneumatic structures address the physical side; lightweight, inspectable computation addresses how data can become a useful signal. Neurytika brings that same emphasis on local processing and clear evidence to enterprise prediction projects.
Celeste’s dAS20 program uses a pressurized textile wing, with aerodynamic lift and electric propulsion. The current prototype has an 8 m wingspan, and the company records its first take-off on 7 May 2026 at Le Havre airport.
Explore the aircraft program ↗Neurytika combines a classification engine with an exploration workflow. Use structured records, build a model, inspect contributing patterns and integrate the result into your organization’s work.
Explore Neurytika software and consulting ↗PUBLIC MILESTONES
Celeste Airships, the company behind the EcoFlyers project, appears among the 2024 laureates announced at VivaTech.
Read the ministry announcement ↗The Ecoflyers dAS10 solution received the Solar Impulse Efficient Solution label on 28 March 2025.
View the labelled solution ↗Celeste’s pneumatic-wing prototype progressed from ground tests to first take-offs at Le Havre, covered by the international technology press.
Read Defence Blog’s coverage ↗Celeste Ecoflyers features in VivaTech’s 2026 press kit through Impact Builders, the program led by VivaTech, Diversidays and L’Escalator.
Read the official VivaTech press kit ↗THE FOUNDERS
CO-FOUNDER & CEO / EMBEDDED AI
PhD in computational neuroscience and author of the Evidence Ratio Classifier research. His work connects interpretable learning, embedded systems and real-world sensing applications.
Read the ERC publication ↗CO-FOUNDER & CTO / AIRCRAFT STRUCTURES
A textile engineer specializing in pneumatic structures and industrialization, bringing materials expertise to the development of Celeste’s inflatable-wing aircraft.
Meet the team on Celeste.aero ↗WHY EMBEDDED AI IS A STRATEGIC BRANCH
For an autonomous drone, processing sensor information onboard can help turn measurements into timely mission signals while making deliberate use of computing power, energy and connectivity.
That is why embedded AI is central to Celeste’s development direction. ERC’s one-pass learning approach and explicit evidence connect hardware-conscious computation with a model that engineers can inspect. The same engine provides the foundation for Neurytika’s explainable prediction workflow.
Telemetry and structured observations
An embedded model close to the data
A category with supporting evidence
Process available measurements on the device and share the resulting category or alert when the mission needs it.
Design the computation around the aircraft’s processing and energy budget, with the learning method and target hardware considered together.
Connect a predicted category to recorded feature combinations and their support, so teams can examine the model’s behavior.
EXAMPLES OF ONBOARD WORKFLOWS TO DEVELOP
Classify structured environmental measurements into an alert category and prioritize observations for an operator to review during a monitoring mission.
Combine recorded vibration, temperature or electrical measurements to identify an operating-state category and provide context for engineering review.
Classify structured observations locally and choose which categories of information are most useful to relay to the ground team.
From onboard sensing to customer, quality and fraud records, the shared objective is clear: a useful prediction, computed in the right environment, with evidence people can examine.
Explore local processing and data control ↗RESEARCH & INTELLECTUAL PROPERTY
ERC development combines peer-reviewed research with patent filings. The scientific paper presents the learning rule, evaluation protocols and results; the patent applications form part of Celeste’s intellectual-property strategy for the technology.
Under examination
Evidence ratio classifier: A one-pass model for fast incremental learning
F. L. Olivier Manette · Neurocomputing, Volume 659, article 131797.
IN THE PRESS
Defence Blog
26 May 2026Reporting on Celeste’s prototype taking off at Le Havre and the engineering behind its pressurized textile wing.
Read the article ↗Clubic
18 May 2026A look at the first tests, the founders and the engineering ecosystem supporting development.
Read the article ↗TechRadar Pro
28 May 2026Coverage of Celeste’s inflatable-wing design and its aerodynamic approach to flight.
Read the article ↗
AEROSPACE + AI + IMPLEMENTATION
Explore Celeste’s aircraft development at Celeste.aero, or talk with our team about a Neurytika software, embedded AI or prediction project.