We have finally managed to get everyone into one place for a team photo. We were missing Etienne Barnard, but we will get him in the next one by all means necessary 😉

RT @ruanvdmerwe1: We have been collecting data to investigate new methods that will allow models to solve visually grounded spoken instruct….

RT @DeepIndaba: Traffic control and RL🚦

RT @ZindiAfrica: Interesting session on introduction to transformers . The process of understanding its architecture is astounding #DLIndab….

RT @dsn_ai_network: Another ongoing session at the @DeepIndaba #DLIndaba2022. Our facilitator is @ruanvdmerwe1, a Machine Learning Enginee….

RT @ruanvdmerwe1: Can not wait for the @DeepIndaba in 🇹🇳 to start!. Super proud to also say I was also involved with the awesome team that….

RT @sumedialab: Join us next week for another exciting Media Lab Monday! Ruan van der Merwe and colleagues are going to be telling us more….

In this paper, our researchers introduce the Representation manifold quality metric (RMQM), which measures the structure of a learned representation manifold. They then show that RMQM correlates positively to the generalisation of neural networks to downstream tasks.

We recently had a paper accepted into the ICML workshop @PTMs_Workshop!. Find the paper here:.

For those who missed #SumoConference2022 or want a refresher, you can find our talk at

Do you love machine learning? Do you love llamas? If you answered yes to any of those questions, head on over to and apply to join our awesome team!.

A very happy birthday to @Mcobus96! May you be able to pet many llamas in the coming year. (Yes an emergency cardboard candle was built by our own Tiffany 😉)

To assess how well these methods generalise, they are tested with both seen and unseen traffic patterns, showcasing how our algorithms outperform previous methods.

We will present how two popular deep reinforcement learning methods can be applied to traffic signal optimisation. This will include a comprehensive review of the behaviour of these different methods and how they compare to current traffic optimisation algorithms.

Join us at #SUMOconference2022 this week.

Our reinforcement learning team is hard at work to create an awesome presentation for the attendees at @EclipseSUMO #SUMOconference2022

If you were at this panel discussion and you are passionate about machine learning, have a look at our Llama (internship) programme at

Last week Tiffany Woodley was involved in a panel discussion, with @predictive_ins and Revolute Systems, on what it means to be involved in the overlapping fields of machine learning and entrepreneurship, hosted by @SULaunchLab and @StellenboschN. Thanks for an awesome event!

For anyone wondering if we love 🦙🦙🦙, we hope our latest art piece provides you with an answer