Blog
Our blog is a place to share our knowledge, talk about ideas and about us. Articles are only available in their original language.
Technology
Reasons to train your own LLM
by Özgür Demir
Most companies use publicly available APIs to incorporate LLMs into their own products. While this is not a bad approach, especially if you're just starting with LLMs, it also has some drawbacks.
Read Post
Technology
Every company needs a modern data architecture
by Özgür Demir
In this video I'll briefly describe on a high level why every company needs a modern data architecture by looking at some industry specific examples.
Read Post
Technology
Can you afford to train AI models?
by Özgür Demir
Training a large language model as massive as Llama 2 70B takes approximately 1,720,320 GPU hours which translates to around 2.5 million Euros.
Read Post
Technology
The Backbone of Custom AI Models: Solid Data Architectures
by Özgür Demir
AI models perform best when trained on problem specific data sets. Data architectures on the other hand allow to ingest, store and process data on a large scale. Thus, without a solid data architecture it is almost impossible to train custom AI models.
Read Post
Process
How to start Data & AI Projects: Prioritize product over technology
by Özgür Demir
The 1st step when starting a new software project should involve understanding the bigger picture of why a certain digital product needs to be built.
Read Post
Technology
How to structure Terraform code
by Özgür Demir
Terraform itself does not specify a specific folder structure. Many engineers are thus confused on how to structure their terraform code.
Read Post