Portrait of Aldric Pierrain

Aldric Pierrain

Founder of Cydral Technology

An engineer for more than twenty-five years, Aldric has carried out applied research and contributed to Dlib for more than twenty of them: images first, from facial recognition to quality control, then language models. With Cydral Technology he publishes this work as a library anyone can read and as products anyone can install.

Experience
More than 25 years of engineering, over 20 of them in applied research with Dlib
Fields
Artificial intelligence, computer vision, cybersecurity, enterprise architecture
Open source
Contributor to Dlib and author of Dlib-Transformer-extensions
Training
Engineering degree in communication systems and networks, INSA Rennes; specialisation in computer vision and distributed systems with IRISA

From image understanding to sovereign AI

Aldric trained as an engineer in communication systems and networks at INSA Rennes and specialised with IRISA in computer vision, image and video processing and distributed systems. His first research covered content-based image retrieval, grid computing and the processing of satellite imagery. Applied research has been his thread ever since: take a method, write it, run it on real data and measure what it is worth.

Images came first. He wrote fastface, a facial recognition engine, and led the development of an image indexing and retrieval platform that qualifies images by their signatures and searches them by their keypoints, then of complete environments for understanding the content of images, usable on an online shop as well as in an industrial process such as quality control. This work has gone hand in hand with Dlib for more than twenty years: he uses the library, extends it and contributes to it, with models for facial recognition, age and gender estimation, image synthesis and colorization.

Language models followed the same path. Transformer layers, a tokenizer and example programs joined Dlib, then the extensions grew into a library of their own. Cydral Technology gathers this work and turns it into products that run on the computers people already own.

His contribution to the initiative

Research and engineering

He takes a method from the idea to the measurement, in C++: computer vision first, then transformer layers, quantized inference kernels, GPU memory extension, training and distillation. Each method is published with its figures.

Product

He takes an idea all the way to an installed program: architecture, installer, signed updates, documentation and an interface in several languages. He has brought software and security products from design to market, hardware included, with teams of developers.

Enterprise integration

He knows how software enters an information system: enterprise architecture, private and hybrid cloud, security requirements, compliance, deployment without a network, and interfaces that existing tools already speak.

Cybersecurity

He builds security into the products from the start: local trust model, encryption, key protection, signed supply chain and documented limits. He works with the reference frameworks: ISO/IEC 27001, NIST, OWASP, GDPR and NIS2.

Trusted AI

He frames the use of AI in organisations: sovereignty requirements, assessment frameworks, operating models and controlled use.

Leadership

He has led teams of architects and experts and steered programmes, from the framing of a need to its delivery.

Publications

Buffer-Level Residency Management for GPU Memory Extension
Aldric Pierrain, Cydral Technology. Research report, September 2026
Technical application architecture of Pseudalis Community
Aldric Pierrain, Cydral Technology. Research report, November 2026
Onward to the Hutter Prize: a token-level transformer compresses enwik9 to 117 MB on one CPU core
Aldric Pierrain, Cydral Technology. Research report, November 2026
BAREL: Backpropagation And Reinforcement-based Environment for Learning
Aldric Pierrain, Cydral Technology. Publication, April 2023
A Small Language Model Stack, Built in C++ on One Machine
Aldric Pierrain, Cydral Technology. Article

Talk with the founder

A project, a partnership or a question about the work: your message reaches him directly.