More relevant marketing content with AI
Date:
Changed on 27/02/2025
Inria Startup Studio mainly incubates business projects led by scientists and engineers involved in one way or another in research conducted at the institute. But the program also welcomes talent from outside the Institute.
A signal and image processing engineer, William Bernal spent 8 years working for big names in marketing, retail and e-commerce, in France and Germany.
Image
Verbatim
At one point, I thought of starting a PhD thesis, but I'm more motivated by designing solutions that people can use in real life.
Auteur
Poste
Porteur du projet Inria StartupStudio Feelim
With an MSc in image and signal processing and a BSc in electronic engineering, he has held project and product management positions in the retail, technology, e-commerce and marketing sectors for over 10 years.
For the past four years, he has worked closely with marketing and content creation teams for social networks. This experience led him to imagine Feelim, which was initially intended to be a content creation platform. After 11 months at Inria Startup Studio, he refined this vision to develop a solution focused on content analysis. His aim is to combine technology and marketing strategy to fully exploit this potential and offer companies concrete tools to optimize their digital presence.
And it's to fulfill this aspiration that he plans to set up his own business. In March 2024, Startup Studio accepted his application for a year's support.
Called Feelim, the project focuses on producing more effective marketing content for social networks.
For marketing managers, it's important to define a good strategy for these media. But social networks can be too complex to understand. You can create content that no one pays attention to because, for example, the post is neither different nor better than the 100 previous ones already published on the same subject. Conversely, if content suddenly goes viral, it's not always clear why.
Based on AI and, more specifically, large agent models, Feelim acts “like a magic wand that helps you understand what's happening on the networks and improve your marketing content. The technology will suggest changes in text and images so that the message becomes more relevant, reaches more people and ultimately results in more contacts. It relies on viral content, influencer opinion and major trends.”
Using computer vision, Feelim also analyzes the images accompanying the text. “Our synthetic eye will indicate that photo B is better than photo A, or that you should change the background, or add a caption to help people understand what you're talking about. If the image shows a bottle, perhaps it's too small in relation to the background. You might want to crop the image so that the bottle is larger. AI will also predict where in the image people will look first. For example, if you have a baby next to diapers, it will say that the eye is on the baby. Now, if you turn the baby's head in the direction of the diapers, then attention will shift to them.”
Feelim takes the form of a web application. You can send it files, or simply a link to a network post. The service offers three options. The freemium plan authorizes a limited number of requests. The subscription plan offers a certain number of 'seats', i.e. analyses of a certain number of networks. And finally, the premium plan is aimed at marketing agencies who need to manage large-scale campaigns on many networks.
A minimum viable product (MVP) was launched in autumn 2024. It gives users “the opportunity to test and receive recommendations. At the same time, we're looking for a company that would be interested in a partnership to gain early access to this new service.”
Visuel
Titre du lecteur
Find out more about Feelim with William Bernal (english)
Fichier audio
Audio file