do machine learning
the right way
Alvíss AI: safe and useful
Our goal is to make safe and useful AI ubiquitous in all industries.
Alvíss is based on a Bayesian Deep Learning technique, only recently available in computing. It combines the power of deep learning with prior knowledge. This enables all models to not only make prediction from knowledge it has learned, but also tells how confident it is about that prediction. Many AI failures you’ve read about can be attributed to this missing piece. To know what they don’t know is absolutely crucial for an AI model to be safe.
In addition to making prediction, Alvíss is also capable of quantifying uncertainty of its prediction. A massive improvement that minimize the risk in using AI.
Our AI is also capable of handling cross domain knowledge transfer and autonomous learning in multiple environments and verticals.
Alvíss AI is also fun to use for a machine learning pro. With ability to configure network architecture, you can tweak the settings to fine-tune your network.
AI As A Service
We streamline workflows of building, deploying and maintaining models thereby reduces time and cost by over 90%. Alvíss AI helps you proceed from development to production phase faster than ever.
With API backend, you can integrate Alvíss AI with your application with little effort. Automate the use of our platform in the way that suits you best.
How it works
You can simply upload your data and start training the model. Alternatively, select a pre-trained models to make prediction straight away.
Click on the diagram to learn more about each step in a typical Alvíss AI workflow.
All capabilities in Alvíss AI, in addition to prediction, quantify certainty of its prediction. This extra measure is the key to improve safety in AI application and prevent your system from acting rashly when it is unsure.
Time Series capabilities takes historical sequential data and forecast the future trend. Many of these problems relate to supply chain optimization, novelty detection, house price prediction, and plenty others that occur on a daily basis for data driven companies.
Instead of forecasting, sometimes you are more interested in what you can do to achieve the best result. Alvíss AI allows you identify target output, and it will predict input values that will optimise the measure. With great flexibility in target specification, you can directly customise it to suits particular questions.
One of the more useful aspects of having a model, beyond predictions, is the ability to evaluate consequences of any deviation from a scenario. With Alvíss AI, you can specify two input scenarios: base scenario, and test scenario reflecting deviation in question. You can then gain an insight of how much each metric may impact the outcome.
Object Detection, as the name suggests, detect and identify object on an image. It enables a number of automation systems that relies on visual cues such as self-driving car, fault detection, etc.
With this example, uncertainty measures from Alvíss AI enables a self-driving car to ask for human intervention when it drives into an unusual scene it has not seen before, and a camera isn’t sure whether it has detected an obstacle or just a part of the new scene.
Image Classification sort images into given categories. One of the areas that benefit a lot from it is medical applications; to detect anomalies in x-ray film or other kind of images where anomalies can vary greatly in its appearance. Image classification can then suggest abnormalities with better coverage than object detection.
In such use case where the result is critical, Alvíss AI provides exactly the uncertainty quantification that you need.
OCR involves reading characters from an image. Alvíss AI can read text in the wild where input is not necessarily document.
Super Resolution takes an image and increase its resolution, allowing you to zoom with quality that is far superior to normal zooming. It can be applied in photography or filming.