Hello world, let's try this again!
This is the official blog of BaseWeight solutions, and this is where we plan to post all information regarding our work as a company, but first we should talk about why we called ourselves BaseWeight in the first place.
The What and the Why of BaseWeight
The term base weight in backpacking is the weight of the pack minus any consumable items, such as water, fuel and food. This is the stuff that doesn't change its weight as you pack it down hundreds, maybe even thousands of miles down a long distance trail. The idea is that the lower the baseweight, the lighter the load and the happier that you will be.
The same is with software. Many existing enterprise ML pipelines are sort of a kitchen sink of dependencies and models, bringing in dependencies that are duplicates of other dependecies and creating this absolue mess where Tensorflow, PyTorch, ONNX Runtime, Mediapipe, OpenCV, TensorRT, Transformers and every other library ever written is sucked into this monstrosity either because of poor architecture, or a complete lack of care by the builder, and the whole system can buckle under the weight of all this stuff, just like a hiker's knees will buckle after carrying fifty pounds of absolute crap that seemed like a good idea up a hill.
The idea of BaseWeight in software is simple, choose what you need, and ONLY what you need. If you need a cross-platform machine learning framework, stick with only that framework. If you need to have a common codebase across mobile and desktop, choose a framework that can support that and figure out how to add the hardware optimizations in that framework. All the ML features should use the same framework whenever possible and things should remain simple.
In the mountains, complexity can kill you, in software development, it just kills your projects and makes you miss your deadlines. Both are bad, and both should be taken seriously.
What we need from you?
Are you a company looking to adopt Generative AI solutions for your business? Do you want to have LLMs with offline functionality in remote places where cell coverage is spotty at best? Are you someone who is just going broke with a huge Amazon bill? We want to hear from you.
Please subscribe to our newsletter, or contact us with your story and we will follow up. We have some unique projects planned for the new year, and the only way we can determine what you want is if you, the potential customer, the user, the person whose success we care about, lets us know about their problems with OnDevice development.
Stay safe out there, and happy trails!