Often, the project stays in a proof of concept and the investments fail to convert into value for the organisation. To speed up the adoption of such AI implementations and avoid project failures, it is possible to use a simplification layer on top of existing- or cloud infrastructure. Using this layer you can easily move from a local algorithm to an operational software application. Individuals and organisations are thus enabled to recoup their data science investments.
What challenge does it solve?
One of the companies offering such software is UbiOps. With their SaaS solution, they simplify the usually complex development of a scalable and robust infrastructure. In this way, parties without (often scarce) IT knowledge of Kubernetes or Docker are able to immediately deploy and maintain a production-ready application.
Accelerating AI deployment
For example, UbiOps works with DuckDuckGoose, a high-tech startup that detects deep fake photos and videos. Together they develop software to identify deep fake images. Deep fake images are made very reliable by the current technological possibilities and can have far-reaching consequences by influencing share prices or elections, for example. The ability to detect deep fake images will play an increasingly important role in the future.
The team of DuckDuckGoose, that mainly consists of data scientists, periodically improves the deep learning model for deep fake detection. It is important to be able to bring new versions of the model into the production environment easily, to perform A/B tests between model versions and also to monitor models. Via the secure API connection with the SaaS environment of UbiOps, the data scientists can do this themselves and are not dependent on scarce IT knowledge. This makes it possible to easily get value out of data-driven applications without IT knowledge.
Collaboration in ethical AI
UbiOps works with various parties on the development of ‘ethical AI’, which includes, for example, the explainability of models, AI audits, but also the ability to detect data drift, perform outlier detection and assess data quality. This is taken up together with other parties in the AI Infra Alliance. The AI Infra Alliance is a group of companies dedicated to bringing together the essential building blocks of the AI applications of today and tomorrow.
Joint ambition
UbiOps: “Through our participation in the NL AIC, we can contribute to the UbiOps: “Through our participation in the NL AIC, we can contribute to the Dutch AI ecosystem and innovate with other parties to make AI deployments a succes. In this way, we help startups in all sectors to accelerate the development of their AI product and contribute to the ambition for the Netherlands to be a leader in AI.”
Interested in more information about the development of deep fake technology? Then also read this article.
