Neuralk-AI is developing AI models specifically designed for structured data

Tabular data is a wide range that includes structural data that usually fits in a specific row and column. It can be an SQL database, a spreadsheet, a .csv file, etc.

Although there has been a lot of progress in artificial intelligence applied to continuous and hierarchical data, these large language models are unclear through design. They are built to manipulate input tokens to generate consistent output without following a specific structure. The best LLMs are either expensive to access through an API or expensive to run in your own cloud infrastructure.

And yet, many companies already have a data strategy with data warehouse or data lake that all important data and some data scientists that can earn this data to improve company techniques.

French startup Neural-II It is working on an artificial intelligence agency by focusing on table data on AI models. This week the agency has announced $ 4 million funds.

“Data with real value for companies that was marked long ago, structured in the form of a table and data scientists used to create all their machine learning algorithms,” Chief Scientist Officer Alexandre PassQ Techchench.

Neural-AI thinks there is an opportunity to reconsider the development of the AI ​​model, but with a specific focus on structural data. At first, it is planned to provide its model as API to the data scientists working for trade agencies because these companies are Love Data – think of product catalog, customer database, trend of shopping cart, etc.

“Today, LLM is great for answering questions based on searching, natural user interaction and structural documents. However, we have some limitations at the moment of returning to classic machine learning, which is really based on the classic tablet data, “PassQ said.

With the help of neural-AI, retailers can automatically automatically automatically automatically automatically with smart duplication and enrichment. However, they can detect fraud, to make the product’s recommendations optimal and sales forecasts that can be used to determine inventory management and product pricing.

The company also led the company’s $ 4 million round with Fly Ventures Stemiye. A number of business Angels also invested in startups, such as hug face to Thomas Wolf, Alan to Charles Gorintin and Philip Carrot and Miracle to Nagi Letifa.

The team is still actively working on its models. It is planned to test with a group of top French retailers and a group of trade startups, such as E.leclerc, Aochan, Miracle and Lucky Cart.

“Within three or four months, we will publish the first version of our model and public benchmark on which we will be able to rank our model compared to the sophistication of this place,” PassQ said. “And in September, the concept must be the best Table Foundation model in all fields related to learning the concept.”