A virtual assistant designed to solve your company’s machine learning problems

As the AI ​​field advances in recent years, more and more companies are developing products based on this technology. However, implementing this successfully requires knowledge of data science and machine learning, so it’s not an easy task. Companies struggle to get the most out of his AI because of the stark professional divide within the industry.

To help businesses, especially small businesses, take advantage of this space and easily build AI solutions, Sapiema built an online assistant named Adam. Assistants act as virtual experts, providing solutions specifically for your machine learning needs. In this installment of the Startup of the Day series, Alexandru-Mihai Glontar dives into solution-based assistants.

What problem are you solving? How are you addressing this problem?

“There is a significant shortage of data scientists and machine learning (ML) professionals in the market. Most within the industry hope that training more people will gradually solve the problem.” However, the reality is that the situation is even worse: the education ecosystem is becoming more and more disparate as people from academia turn to industry for higher salaries.

So we are virtually automating the development of AI-based projects by turning anyone, regardless of technical skill, into a machine learning specialist. We turn them into data scientists by giving them the right tools: assistants who do it instead of data scientists. ”

How does this virtual assistant work?

“This is what we call virtual machine learning engineers, with a conversational interface that gives users the same experience as a remote engineer sitting anywhere in the world. Open a chat and have a conversation with this assistant, a virtual engineer, who will explain the problem at hand in terms of business requirements, and then a virtual engineer named Adam will guide you through the entire process of gathering relevant information. Let’s start to guide you.Conversations are designed to be accessible to everyone regardless of their level of technical knowledge.Generally what we expect from our users is that they want Adam to deal with It’s about having knowledge of the problem, and during this conversation Adam just tries to fill in a configuration file, based on which the whole solution is generated, automatically trained and deployed on the customer’s infrastructure, It just outputs a web API that customers can integrate with the product.”

Are there potential data privacy issues with large data transmissions?

“The entire training, data transformation, data cleaning part and model deployment happens on the customer’s infrastructure. Even if we knew, we would never access the actual data.”

How are innovations different from existing ones?

“We are in a very hot space and we have very large companies as competitors. , almost all of which focus on making data scientists more productive, and the fact is that for many companies, especially small and medium-sized businesses, having a single data scientist is nearly impossible. Yes, what we do is help non-technical people manage such projects on their own: data engineers and backend engineers who want to work on machine learning projects. Yes, but I don’t have the time or budget to learn about machine learning, so tools like this help me get started quickly without a steep learning curve.

There are ML cloud providers focused on developing large models suitable for wider scale. For example, if you look at Google Cloud, there are text models, image models, and video models, but the reality is that these kinds of models are too general to run effectively. Our approach is therefore to train a small model for a specific problem, a specific dataset. This can ensure that predictions and classifications are correct most of the time, even if they are a little less general.

And most importantly, we are the only solution provider on the market that offers a fully guided experience. No other provider offers such a fully supported experience, making it very easy to adopt our solution as there is no learning curve. Get more done by chatting like you would with a colleague. ”

What’s the biggest obstacle you’ve faced so far?

“What we offer is a bit sci-fi, so building trust with our customers has been our biggest challenge. And we are still developing the first version of the product. So even if we show you a demo where you can already train a model and deploy it, the customer feels that the end goal is technically unattainable, and it takes a lot of background to understand. It’s hard to understand how we get from point A to point B. So people’s immediate reaction is, “Okay, if you don’t understand us, point B may not be reached.” But we know it is possible.

We are already very close to launching the first version of our product. Resolved technical issues encountered. At this time, we need to build a better communication strategy to convince potential customers of our product’s capabilities. ”

What does the future hold for Sapiema?

“First of all, we need to finish the pre-seed round, but it is progressing very slowly due to the turmoil in the global economy. Success is even more important, our main goal next year is to become an AI solution provider in the European fintech market, although we also want to work on healthtech and legaltech. , fintech is a good use case for our technology.Large financial institutions have teams of data scientists who innovate at a fast pace, so SMEs need to keep up with the pace of innovation, and here I am. We will intervene.”