AI-Stack

100% Open-Source and 100% free AI Stack

 

Vanilla AI-Stack consists of 100% Open-Source components. It can run everywhere where a GPU or NPU would be available - be it Data-Centre, Public Cloud or Private Cloud. It can run any AI-Model and Machine-Learning Workload from Huggingface or other marketplaces. It does not incur any licensing fees or token-based costs.

 

 

As AI-components can be expensive and hard to purchase, Vanilla AI-Stack runs off commodity GPUs, as long as they support CUDA or ROCm. Often, the vendors force customers to purchase extremly expensive middleware to facilitate GPUs or NPUs across multiple workloads. This is mitigated with Vanilla AI-Stack, as we use Kubernetes as underpinning, allowing for time-slicing or parallelism with GPUs.

We have developed the Knowledge Pyramid, an approach allowing to structure knowledge for Generative AI-models into different layers of priority. This allows for more intelligent, more customer- and use-case-focused answers, without enforcing the need to specifically train models, which is utterly expensive and requires a lot of time to be spent.

 

 

Foundational Knowledge

Foundational Knowlegde is the base layer of knowledge. It typically contains general knowledge, such as best-practices, blue-prints, approaches, general regulations. This layer of knowledge shapes the use of the AI stack.

 

Specific Knowledge

Specific Knowlegde adds regional and industry-vertical specific knowledge. It builds upon the foundational knowledge, and adds information to it. This allows to fine-tune the stack towards a certain environment and use-cases.

 

Custom Knowledge

The final layer is Custom Knowledge. Here, customer-specific knowledge, their own regulations, use-cases, and information is added. This allows to customize the stack towards its specific usage-scenario, without having to re-train it.

Vanilla AI-Stack supports many use-cases. The following is just a very small excerpt of what is possible. Reach out if you want to learn more or discuss your own use-case!

ESG Assistant

Custom Solution based upon Vanilla AI-Stack to process ESG Regulations, a diverse range of data sources (including Word-Documents and PDFs), create Checklists, create ESG-Report Drafts, and to query the data using a Chat-frontend.

Hospitality Solution

Custom Solution utilizing Generative AI and Machine Learning to predict bookings, occupations, prices, and to correlate them with each other. Create Offers and analyze competitor websites. Performs Benchmarking on data.

Document Management

Custom Solution to perform OCR on scanned / sent documents regardless of their format, to analyze their content, to assign them to different document- and object-types, and to trigger additional workflows based upon content and state.

We support our customers in setting up, running, and maintaining Vanilla AI-Stack, as well as developing solutions based upon Vanilla AI-Stack:

• Architecture
• Solution Design
• Automation
• Roll-Out
• Integration of Hardware
• Operations
• Maintenance
• Software Architecture
• Software Development
• Scaling and Resilience

Our Architects, Engineers, and Operations Teams are experts in Vanilla AI-Stack - on any On-Premises-, Private Cloud- or Public Cloud-environment! Reach out to us to discuss your requirements!