AiThority.com Interview with Nicolas Gaude, CTO at Prevision.io

Hi Nicolas, welcome to AiThority Interview Series. Please tell us a little bit about your role at Prevision.io and how you arrived here?

My name is Nicolas Gaude. I am a physical engineer by training as well as the co-founder and CTO of Prevision.io. With 20 years of experience in the field of software development and as a Big Data and Data Science expert, I’ve led the R&D of large companies in these fields (La Poste, Bouygues Telecom, NDS / CISCO). I’m also humbled to have been recognized as one of the best data scientists in France following numerous victories in dedicated competitions (Kaggle.com, MeilleurDataScientistDeFrance, DataScienceOlympics, DataScience.net).

What is Prevision.io and what are your main offerings?

Founded in 2016 by a team of renowned data scientists, Prevision.io brings powerful AI management capabilities to data science users so more AI projects make it into production and stay in production.

Prevision.io’s purpose-built AI Management Platform was designed by data scientists for data scientists and developers to scale their value, domain expertise, and impact.  From banking and financial services to healthcare and retail, data scientists too often lack the tools to create efficient data models.

Which industries and regions are you currently offering your AI services to?

With Prevision.io, a member of the Google Cloud Partner Advantage program, data scientists and analysts have the tools they need in one place to build, deploy, monitor, and manage data models across a variety of industries. However, the platform can be accessed in most places in the world.

AI-as-a-service is one of the biggest technology markets in the world. How does Prevision.ai serve this particular segment with its robust AI platform?

We bet that this market is now mature enough to welcome autonomous solutions that target non-AI experts. The Prevision.io platform is designed to offer citizen data scientists robust and efficient tools to deploy and maintain complex predictive models with no additional effort.

Tell us more about the engine driving your AI management platform? What kind of infrastructure does a company require to fully benefit from Prevision.io’s  AI and ML offerings?

The beauty of SaaS is that there is no requirement for a company to adopt the Prevision.io platform per se. If you subscribe to Prevision.io, the engine will deploy all the cumbersome technologies and infrastructure. There is no need to have a Kubernetes cluster, to handle complex notebooks, or to operate data pipelines. We set up the service for you.

AIops and AutoML– how do you differentiate between the two technology curves at Prevision.io? Could you elaborate on these two AI branches for our audience?

At the heart of the Prevision.io, there is a powerful autoML engine. This is indeed the core feature of an AI platform. But with that said, we also provide the shell around this core with a scalable and resilient infrastructure, namely AIOps, as we understood that this is the turn-key and also a pain point for companies to put AI in production. AIOPs and autoML seamlessly combined into the same platform gives our prospective clients a compelling reason to go with Prevision.io compared to the painful integration of distinct systems.

How do you see the competition in the AI market further disrupting the need for Open Source data and DevOps communities?

Open source data and open source infrastructure are two important pillars for AI tech companies such as Prevision.io. The competition is fierce for open source data. For example, Dark Sk is a weather data API that was acquired by Apple, and its service eventually shut down. Kaggle, acquired a long time ago by Google, is now aggregating an impressive collection of datasets. In the DevOps field, open-source is still king, Kubernetes reigns supreme, and cooperation between open source actors brings a shared value greater than that which could be obtained by a single actor.

What does your product and engineering roadmap look like for the years 2022-2023?

With our AI management platform live and running, all our efforts are now geared towards the downstream of AI. The downstream of an AI is all about deployment, scalability, maintenance, and monitoring of predictive models. The next step for us is to provide—alongside our AI platform,—an AI marketplace to enable our customers to exhibit and monetize their AI applications based on the models they build inside the Prevision.io AI platform.

Which technologies are influencing the recent innovations in the modern Cloud and Infrastructure industry? How do you see the rise of emerging techniques in AutoML and AIops benefitting the whole IT industry?

For years now, Cloud innovations have been related to containerization. This is where the Kubernetes ecosystem delivers a large part of the innovation. Beyond this, companies are developing tools to decouple their application from a specific Cloud vendor, and still benefit from Cloud vendor managed services. In this area infrastructure-as-code is the way to go, whether you choose Terraform or pulumi, both bring a unified API to deploy your cluster with attached managed services on top of different Cloud providers. This lean-and-shift strategy enables companies to move rapidly from one configuration to another (from the public Cloud to a private Cloud, or from one vendor to another).

Your forecast for the skills and talent required to excel in data science world: What would be your advice to all the young AI engineers and data science analysts eyeing an opportunity in companies that belong to Prevision.io’s league:

When we started Prevision.io five years ago, it was all about data science and skills in algebra, statistics, and development. Now we see a much more rationalized knowledge tree balanced between data and IT talents. We still need expertise in data science talent to understand, test, and validate new ML techniques, algorithms, and components, but the major part of the work is now about integration, deployment, and wrapping API and user interface of these ML features.

So my advice, if you’re a young AI engineer, is to compliment your experience with dev-ops to understand the production challenge of AI: from scalability to resilience to monitoring. Or for dev-web to understand the usability challenge of AI, including user-interaction, api-design, and outcome-visualization.

Your advice to your C-suite peers in the industry on how to make AI ethics central to any organization’s digital transformation:

Data professionals should strive to make purely objective assessments while communicating or documenting results and performance metrics, and sharing their recommendations of the technical methods. Too often, it’s easy to jump to conclusions from what the data scientist thinks the data says rather than just presenting others with the numbers themself. The data professional should also never make false, exaggerated, or misleading claims concerning the performance and efficacy of a system. So when it comes to AI analytics, the safest route is to let the program run its course without interference, as this adds to the possibility that there will be conscious or unconscious bias integrated into the interpretation of the results. The data professional should stick to documenting and communicating the objective facts rather than inserting themselves too much.

An AI domain that you are keenly following and hope to be part of in future:

Quantum machine learning, while still in its infancy, is a very active topic with interesting promises in terms of resource efficiency to learn models with a dramatically huge number of features.

Thank you, Nicolas! That was fun and we hope to see you back on AiThority.com soon.