Parisian startup Pathway, which builds a technology to feed AI models with real-time, changing data, has raised a $10m seed round — just as it moved its headquarters to the US.
It comes two years after the company, which has already landed NATO and the French postal service as customers, raised a $4.5m pre-seed round. The new fundraise was led by US investor TQ Ventures, with participation from Swiss VC Kadmos Capital, Polish investors Inovo VC and Market One Capital, and French investor id4 Ventures.
A “large majority” of the fundraise was equity funding, according to the company, but it also included some debt.
The startup was previously backed by angel investor and OpenAI researcher Lukasz Kaiser, who co-invented the Transformer architecture behind ChatGPT. Kaiser is currently an advisor to the company.
Moving to Menlo Park
Zuzanna Stamirowska founded Pathway in 2020 after developing, as part of her PHD, an AI model to forecast maritime trade. She teamed up with CTO Jan Chorowski, who was previously a researcher at Google and Microsoft, as well as COO Claire Nouet, an alumni of French universities HEC and Sciences Po; and CPO Adrian Kosowski, a computer scientist from French national research institute Inria and engineering school Polytechnique.
The company has just moved its headquarters to Menlo Park in the US, where Stamirowska is also preparing to move. She says that most of the R&D team will be staying in France, but that it made sense commercially to double down on the startup’s presence across the Atlantic.
“We are keeping an entity in France,” says Stamirowska, “but as we develop commercially we need to be where our users and customers are.”
Pathway’s customers include French organisations like logistics multinational CMA CGM and public transport company Transdev, as well as US-based enterprises like tech giant Intel; the company says that it is seeing most traction among the developer community in the US.
“We need to be in the room where it happens — and it happens in the Bay Area,” says Stamirowska.
What does Pathway do?
Current AI models, including Gen AI models, are trained on static datasets — meaning they are not immediately reactive to new information. In industries like logistics, where data is constantly changing, this can be a significant limitation, says Stamirowska.
“LLMs and machine learning models in general are trained on a dataset, and the way to treat new information is through retrieval augmented generation (RAG) — essentially a big data pipeline for engineers to put changing data into the LLM and then make it respond,” Stamirowska tells Sifted.
“Our technology allows them to deal with live data, which they can feed in very dynamically into the LLM.”
Pathway provides a framework that developers and engineers can use to feed real-time data to AI models. The startup says this enables those models to make decisions based on up-to-date information.
The technology is relevant for AI and machine learning models in general, but the company says that it has seen particular traction for Gen AI applications in recent months.
From NATO to Formula 1
Pathway’s framework is open-source, meaning that developers can download the code for free and access its core features.
Enterprises which wish to use the framework for commercial purposes, however, have to license the framework. The startup is already making revenue, says Stamirowska, although she declined to specify how much.
In just five years, Pathway has secured some high-profile customers. Last October, the startup announced that it was working with NATO to support data processing from military sources, as well as open-source information such as social media alerts and civil traffic.
French postal service La Poste also uses the platform to map its transport processes, which the startup says enables the organisation to supervise and optimise its operations. The technology proved particularly useful, says Stamirowska, during the Summer 2024 Olympic Games in Paris, when there were more traffic disruptions and road closures than usual.
The startup also worked with a Formula 1 racing team to manage the real-time processing of millions of data points during races. “A racing car produces masses of data,” says Stamirowska.
“Every car is a prototype, so engineers are pushing the hardware to the limit to get the best performance. And then of course there is the race strategy.”
“It’s a great tool for automotive engineers to build their own dashboards to monitor and supervise the race.”
Pathway plans to invest mostly in R&D in the coming years to work towards what Stamirowska describes as ‘live AI’ — making AI models “more responsive and reactive,” she says, by enabling them to put information into context.
The startup is recruiting for a dozen roles across all departments (R&D, sales, marketing, engineering, product) both in the US and Europe; and is planning to double down on the US for commercial growth.
Read the orginal article: https://sifted.eu/articles/pathway-10m-seed-round-news/
