Keywords: Deep Learning, Artificial Intelligence, Speech Recognition, Natural Language Processing
CONTEXT
LINAGORA is a world leader in the development of innovative open-source solutions. By sharing its open-source software, LINAGORA aims to fulfill its mission of providing universal access to the best digital technology available and helping its customers attain digital sovereignty by freeing them from global software giants and proprietary platforms.
Based in France, LINAGORA has additional offices in Canada, Vietnam, Russia and Tunisia. While it shares its software all over the world, LINAGORA focuses its efforts on developing countries, especially in Africa.
LINAGORA is a rapidly-growing, innovative company of nearly 200 employees whose founders control 100% of the capital. Created in 2000, it is now a French leader in open-source software development, a position for which it has received numerous awards and prizes. It is a member of the Cap Digital competitiveness cluster as well as a partner of the Interdisciplinary Institute for Artificial Intelligence (3IA) ANITI in Toulouse. It is also firmly committed to the French Tech movement.
OBJECTIVES
LINAGORA (http://linagora.com) is currently developing the LinTO toolbox, which includes a variety of open-source components to add voice control to applications.
LinTO is designed to offer assistance during meetings (smart meetings), where it can be used to help participants by operating equipment in the meeting room (lights, projectors), making relevant recommendations, transcribing meetings and even generating meeting minutes. However it can assist in a variety of other contexts as well: "busy hands" assistance, voice control of web applications, transcription APIs, subtitling of web conferences, etc.
As a part of the LinTO platform, we have developed a wide range of open source technologies dedicated to speech, including speech recognition and adaptation of wake words, synchronous and asynchronous transcription, editing of transcribed conversations and post-processing.
Our automatic speech recognition engine for French, which is based on the Kaldi toolkit, achieves state-of-the-art performance, surpassing most of our competitors in the transcription of spontaneous speech (cf.https://linto.ai/enterprises). To push this performance even further, we are currently working on end-to-end Deep Learning approaches (deep neural networks) as well.
We are looking for an engineer/researcher motivated to help us advance on these topics.
RESPONSIBILITIES
The selected candidate will join LINAGORA Labs (https://labs.linagora.com) as part of a multidisciplinary research team with a strong Artificial Intelligence component. Their work will focus on the development of a transcription engine for natural language based on an end-to-end Deep Learning approach.
They will also participate in the implementation of the transcription engine in collaborative R&D projects as well as in customer contexts.
The results of the work will be placed under an open license and shared with the community.
CANDIDATE PROFILE
We are looking for candidates with the equivalent of a PhD in Computer Science (BAC+8) with a focus on speech recognition. The candidate must have strong programming skills (Python, C++) and a mastery of Deep Learning methods.
They should be ready to invest themselves in all aspects of development of the transcription engine pipeline and manifest a desire to evolve in an innovative, technological environment.
The candidate should be highly motivated, rigorous and autonomous, with a thirst for discovery and experimentation and the capacity to bring forward new proposals and defend their technical choices. They should moreover share LINAGORA’s commitment to free software and contribution to user communities (lists, FAQs, HOWTO)
How to apply
Please send your cv to [email protected]