Introducing SnowwAI

Introduction

We want to enable European companies to transform with AI in a more customized, transparent, sustainable and cost-efficient manner, without tethering them to us or our cloud forever.

Motivation

As Generative AI is getting more and more capable and takes more and more space in our daily work and as our dependence on large cloud providers and the risk they present becomes more apparent, we would like to propose an alternative that:

Is based on open-source and open-weights models

The open-source/open-weights model space has been catching up with commercial models and is now activiely completing with them. The open source community is very strong in this area and keeps on delivering new and improved models every few weeks.

You remain owner of your data:

We are not in the AI model race, we are here to provide you a platform to transform your company around AI. We will use your data how you tell us to use it.

Has no vendor lock-in:

We are not suggesting that you tether your company forever with our platform instead of an other. We understand that this would replacing a vendor lock-in with an other. We will avoid that by:

• Using open-source/open-weights models as basis: We are using open-source/open-weights models not only because they are more competitive, allow us to control our data and make it more sustainable, but also because it allows you to build your own infrastructure with those some models if you need to do so in the future.
• Your data, your models: If we train models for you, you will have the ability to download them and run them on your infrastructure if you choose to do so. That way, you do not need to re-invest in model training or application rebuild just because you are changing providers.

Is more sustainable than others:

We are not claiming full sustainability of our offerings. What we commit to instead is:

• Lower energy use: Limit our energy consumption by recommending the use of smaller models over larger ones. Larger models will drastically consume more power than smaller ones, while smaller models can be "good enough" for more tasks than we might first assume.
• Longer hardware lifecycles: Extend the lifetime of our hardware beyond the typical depreciation time. Hardware has a useful life beyond the typical 3-5 years and its recycling is more and more complex and polluting.
• Avoid training LLMs from scratch at all costs: Sometimes, standard models will not be good enough or sufficient to fulfill a task that you want to automate with AI. If we are in a position where we can train models, we will never to do it "from scratch" as it is extermely resource intensive. We will always use an other model as the starting point to fine-tune for the task in question, instead of re-inventing the wheel for brownie points.

Has transparent pricing:

Some providers hide their price list behind "Contact Us" buttons. We'd like to be as transparent as possible when it comes to our products

Isn't crazy to compete in the cloud?

This transformation is so deep we can't afford not to compete

What do you offer?

Currently, we have deployed 5 Language Models of various size and capabilities, 4 Embedding Models and 1 Guardian model. See more details here:

Can I train or fine-tune my custom models?

This is definitely I want to offer, the code and the infrastructure of SnowwAI is not there yet. Feel free to contact me if the standard models are not fulfilling tasks they should be able to, and let's see together if there is a way if fine-tuning might be the solution!

How do you get started?

Right now, you can subscribe to our LLM-as-a-Service API. You get a key that allows you to access a standard API (compatible with OpenAI standards), you'll be charged monthly based on your consumption.