Training for 2027

Running local models by yourself? We can all do that now, and having your own Nvidia infrastructure has become commonplace in many companies. With this, you not only have the hardware for inference but also everything needed to train models. While you won't create the model that OpenAI or Anthropic is still missing, there are numerous examples where it’s worth it – such as your own Dense/Sparse/ColBert embeddings, small specialized models for data extraction and preparation, and much more. The best part? By 2027, you won’t have to be clever, because a new colleague not only knows better but also has more endurance: the agent in the AutoResearch loop. So, if you want to know today how to train models by yourself in 2027 – feel free to come by and bring ideas, experiences, and notebooks for Colab… our speaker (and CTO) Johann will appreciate it if some of you join in. --- Doors open at 5:30 PM, food will be provided. Note: Not a beginner-friendly topic!