Runfish-data teaches AI interessants in 1-on-1 sessions and also offers group teaching, in accordance to customized clients need.
To help you or your business, we like to give insights into the working dogma that we use ourselves. It’s when and how to use the abundant tools, such as chat AI and search AI, which make 80% of the difference. At runfish-data, we use Prompt engineering roughly 1 hour per day, and hence developed a whole system of when to use which prompting strategy and crucially, when to give up, because recognizing that even 100 iterations make no difference, is still very relevant, as AI is no paracema (yet).
For the folks who want to know the mechanisms that enable AI today. This course is about Artificial Neural Nets, which are the driving force behind today’s AI revolution. The technology found practical applications early in finance but the example of handwritten digit recognition is the most popular one , MNIST, a dataset of 128×128 binary images, was the most popular AI dataset for a long time. In 2014, it was ousted by a model that identified cats and dogs on images, and AlexNet as it’s called, was recognized as real breakthrough. However, nothing would be possible without the Neural Net Architecture that was first proved colossal in 2018, the ‘transformer’-block, which is found as a building block in probably all leading foundation language models.
Forget about Siri or other virtual assistant. In 2025, The chat AI planning capabilities have cumilated to finally give reliable ToDo lists, which they autonomously walk through step-by-step. Here you learn how to set this up so your virtual assistants can perform tasks autonomously
Today, the cumbersome part of coding is history. I rarely write code, but I delegate it predominantly to chat models. I don’t read it but simply copy pasting.
what remains is reading code and debugging.
We distinguish ourself from other AI trainers due to being having the most recent information. Because of untraceable fast developments of AI models, where a state-of-the-art model can predominantly only remain top for ~1-4 months. Therefore, to stay up-to-date, developers are under pressure to spend a lot of time, which they often don’t spend, because it’s exhausting. We do spend the time and go down all the top-relevant rabbit holes.