TVM Conference Visual Identity
As computers become more powerful, and access to data increases, there is a need to bring greater intelligence to our electronic devices. The ability to run sophisticated deep learning models—those embedded within the technology we use on a daily basis—opens up many possibilities. Auto-correcting cameras in smart phones, object detection algorithms in autonomous vehicles, and speech recognition software in home assistants (e.g. Amazon’s Alexa) are but a few examples. TVM, or Tensor Virtual Machine, is an open-source deep learning framework that bridges the gap between machine learning models and launching them into production. The annual TVM conference, known as tvmcon, is an opportunity for scholars, engineers, users, and practitioners to come together and present case studies, industry advancements, and new directions in research. The hope is that this knowledge can be leveraged to push the limits of discovery, amplify the reach of TVM software development, and raise the collective performance of machine learning.
Now in its 4th year, OctoML, the event organizer, was looking to reimagine the event graphics through the introduction of a new logo, visual language, and website. The concept I presented focused on developing a modular design system featuring “tiles” inspired by the drawings and diagrams that help explain what deep learning is, and how it works (see above). These include various algorithms, pattern recognition, neural networks, and multidimensional tensors. As the system was applied across key touch points, from media promotions to event collateral to merchandise such as t-shirts and tote bags, multiple layout configurations emerged ranging from simple to complex, and playful to sophisticated. While the creative parameters and production specifications for each item were all unique, a modular approach to the visual identity gave us the flexibility to solve a variety of design challenges differently, while keeping everything familial and cohesive.
