Data integration is a necessary part of many workflows from onboarding customer data to exercising payroll, but for many data sets the process is long and manual. Data is siloed into databases and SaaS applications that each keep the information in different formats, which makes it difficult to move information from one database to another. Lume looks to change that using AI.
Lume’s system uses AI and algorithms to automate data mapping by extracting data from its database silos and “normalizing” it, meaning it converts it into a standardized format so that it can be more easily moved or integrated into other workflows. If a data integration breaks in the process, a common problem, Lume notifies users and uses AI to try to remediate the issue. Lume also offers an API and a web platform so clients can embed Lume directly into their workflows.
What sets Loom apart from other data mapping tools is that it isn’t focused on extracting data from a spreadsheet or PDF but rather on complex nested data formats like JSON. Lume can help companies map complex arithmetic, taxonomy and text manipulation tasks, Nicolas Machado, Lume co-founder and CEO, told TechCrunch. He added that this focus allows companies to spend less time and money than outsourcing these data projects.
“One of the core problems that we saw is that moving data seamlessly, like truly seamlessly between systems, is a completely manual process, and has been for literally 60 years,” Machado said. “Why can’t this be automated? Why was it never possible before? It’s because data is unique to every single system. Each company, each vendor, each integration, they’re defining data in their own way. They’re structuring data in their own way. They understand the data differently.”
Machado and his co-founders, Robert Ross and Nebyou Zewde, know the problem well. The trio met as freshmen at Stanford, studying computer science with a focus on AI. From there, they went off to work at tech companies ranging from Apple to OpenDoor, they all were involved in data integration projects. In 2022, when the founders saw the writing on the wall for advancements in AI, they decided it was the right time to try to solve this data integration issue.
“Every engineer has faced this problem,” Machado said. “Every engineer has to do it. So we got started, we got together, literally went to one of my co-founder Robert’s apartment, and we would work the evenings on it.”
Lume was founded in January 2023, launched its first product in March 2023, and went through Y Combinator’s W23 batch. Machado said the company has had strong inbound demand ever since and has racked up dozens of customers so far. Machado said Lume’s customers range in size from startups to Fortune 500, but declined to share specifics.
Lume recently raised a $4.2 million seed round led by General Catalyst with participation from Khosla Ventures, Floodgate, and Y Combinator in addition to angel investors.
“They really understand this problem, and that’s why they were hooked on it,” Machado said. “That’s why they’re excited about it. They were operators. Like, ‘wait, I was a CEO 30 years ago, and this was a problem. This still a problem? That’s insane.’”
Machado said that the round will be used for hiring, the company wants to double its headcount from five to ten by the beginning of next year, and for continuing to work on the tech.
Lume isn’t the only company looking to fix data integration woes. SnapLogic is one that has raised $371 million in venture funding. Osmos is another startup looking to help companies with this. As engineers continue to deal with this issue, competition will likely grow. Machado said he isn’t worried about competition though as he thinks Lume’s algorithms, and how its API brings Lume into company’s existing workflows, will help it stand apart.
In the future, Lume hopes to be the glue that sits between any two data systems and can seamlessly facilitate the flow of data between the two.
“We all love data, and we’re all big believers in how important data is,” Machado said. “That metaphor we use, it’s like oil, oil, historically, to extract value you have to process it and then use it to power machines, everything else. That’s exactly what data is and what data has been.”