Luminovo builds tailored AI solutions and software to augment human intelligence. Luminovo's hybrid learning platform helps to automate repetitive workflows based on images and text as content screening and document mining. This allows for higher quality outcomes at a lower cost by having humans focus on the most difficult cases and letting continuously improving deep learning models do the rest.
Typically, we train deep learning models based on data supplied by our customers in combination with state-of-the-art transfer learning techniques. If we have to manually label data, we do so by using smart labeling interfaces powered by active learning to reduce the labeling overhead and reach a higher performance more quickly. For the deployed models we often use our hybrid learning setup, where the model can query the human in uncertain situation. This feedback is then stored and used for re-training purposes to improve performance over time. To give these processes a head start, we have assembled an arsenal of pre-trained models and an integration for cloud APIs of GCP and AWS. However, we believe that out of the box models are OK for everyone but good for no one. The main purpose of our hybrid platform is to create a sustainable process which builds your custom model on the fly.
At Luminovo, we are all deep learning natives. Our current team is made up of best-in-class engineers from Stanford University, ETH Zurich, TUM, and CDTM. After stints at Google, Intel, and McKinsey, our two founders met at Stanford University and decided to relocate to Munich in 2017 to help European businesses accelerate the adoption of deep learning. So far, our clients include startups from Silicon Valley, mid-sized German companies, as well as established DAX corporations as Linde, ProSiebenSat1, Audi, Infineon and Munich Re.