We present an
initiative that brings together industrial and academic research to achieve common goals, taking advantage of the synergies of both worlds.
The
Digital Life Disruption Lab team of the Discovery unit in the CDO area os Telefónica Innovacion Digital focuses on the innovation of digital products and the study of disruptive technologies such as Artificial Intelligence and quantum technologies, seeking to ensure their accessibility for all. This team works on two fronts:
creating technological experiments to validate hypotheses on the adoption of digital products and the impact of emerging technologies, and
disseminating the knowledge acquired to other teams and units. In addition, it participates in the
training of interns at Telefónica and proposes challenges for the TUTOR Programme, strengthening ties with the academic field.
The
TUTOR Programme aims to
bring undergraduate or master's degree students closer to the company in the final phase of their studies, proposing real market and industry challenges that students develop together with professional experts for their final degree or master's degree projects. This benefits both the students, who are looking to excel in their final projects, and the Digital Life team, which is researching in areas such as Artificial Intelligence, agents and quantum technologies.
A
success story of the program is the joint work of Fernando and Lidia. Lidia, with studies in biotechnology and a Master in Data Science, and Fernando, a Telecommunications engineer and Master in IoT, specialized in audio processing and doing an industrial PhD,
collaborated on a project to improve interaction and privacy through deep learning and Edge AI, eliminating noise that affects speech quality.
The results were outstanding: Lidia obtained honors in her Master's Final Project and, together with Fernando,
they published their work at the Iberspeech 2024 conference in Aveiro, Portugal. Their publication presented MiniGAN, a new generative adversarial neural network designed to improve speech signals in real time, while maintaining efficiency and competitive performance. MiniGAN uses a simplified encoder-decoder structure with conformer blocks and residual connections, which reduces its computational complexity, making it suitable for devices with limited resources. You can see more details here and also a demo.
This initiative demonstrates how
collaboration between industrial and academic research can lead to significant advances in technology and its practical application.