Extrality

Extrality is a Paris, France-based software company founded in 2019 that develops a platform for rapid numerical simulation using machine learning [1][5][6]. Its stated purpose is to bring deep learning technology into engineering design offices through a simulation platform built on neural networks [6]. According to a company-sourced profile, the platform lets engineers assess the aerodynamic, fluid and thermal behaviour of a product in seconds while retaining the physical accuracy required for industrial use, with four patents pending at the time of the description [4].

The company positions its technology as a way for manufacturers in aerospace, transport and energy to reduce time-to-market, freeing engineers from long and tedious computations so they can focus on product development [4]. French registry data classifies Extrality under IT and web development, with a B2B model and activity described as a web platform/mobile app service [5]. Its registered legal form is a société par actions simplifiée (SIRET 85330631400016, NAF code 6311Z), registered at 84 Rue du Faubourg Saint-Martin, 75010 Paris [5].

Extrality has also published open-source research artefacts through its GitHub organisation (now titled "SimAI"), including AirfRANS — training scripts and a dataset of incompressible Reynolds-Averaged Navier–Stokes solutions over NACA 4- and 5-digit airfoils presented at the NeurIPS 2022 Datasets and Benchmarks Track — as well as related simulation and infrastructure repositories [2].

Founding story

Extrality was founded in August 2019 by Nicolas Rasamimanana (CEO) and Pierre Yser (CTO), who hold doctorates in machine learning (time series) and computational fluid dynamics respectively [5]. The initial idea came from Pierre, who during five years at Dassault Aviation as a numerical simulation research engineer experienced how long and tedious computations constrained engineers' inventive work [5]. He met Nicolas, an entrepreneur and academic researcher who had spent six years as sole founder of a technology startup and seven years in an academic machine learning career examining the links between physical and statistical models; Nicolas was a member of the Entrepreneur First EFPA2 cohort from April to September 2019 [5]. A separate profile describes the founding pair as an engineer specialised in mechanics for aeronautics and space and an entrepreneur with innovation expertise [6].

Business model

B2B software, per French registry classification, delivered as a web platform/service to industrial manufacturers [5].

Traction

A company-sourced profile describes a team of 15 people combining machine learning and physics expertise [4]; French registry data places headcount in the 11–50 band [5]. Public research traction includes the AirfRANS dataset and accompanying training scripts accepted at the NeurIPS 2022 Datasets and Benchmarks Track, with the corresponding GitHub repository holding 69 stars [2].

Latest developments

The extrality.ai domain now serves an Ansys/Synopsys "Ansys AI" page covering AI-augmented simulation products including Ansys SimAI, GeomAI, TwinAI and Engineering Copilot, and references an Ansys 2026 R1 release [0]. The company's GitHub organisation is titled "SimAI" and shows repository updates through 2026 [2]. Separately, the French startup registry myfrenchstartup lists EXTRALITY as struck off ("radiée") since April 2024, with the last BODACC filing recorded as a share capital increase in December 2022 [5]. No source in the material explicitly states an acquisition, merger or shutdown.