Ryan here, one of the co-founders of Lium.
A quick story on how we got here.
A few years ago, my co-founder and I had been working with AI long enough to see both its immense potential and its limitations. The part that kept bothering us was this: so much of the world’s most important data is too large, too complex, too multimodal, or too domain-specific for today’s AI systems to work with well.
And we kept coming back to the same question:
What kind of positive impact could AI have if it could actually reason over the kinds of complex data that power science, climate, energy, advanced manufacturing, and other critical industries? Our team is deeply impact-driven, and our lead investor SJF Ventures is an impact fund, so we set out to solve this problem.
To really test the limits of AI, we decided to start with one of the hardest physical data domains we could find: astrophysics.
We collaborated with astrophysicists who specialize in interpreting Chandra data. For context, Chandra is NASA’s great observatory for the X-ray band of the spectrum. Our north star was simple: could we make it possible for someone to ask questions of incredibly complex scientific data in natural language? Could a 12-year-old have a conversation with real data from a black hole?
Over roughly a year of research, we tested a range of AI and machine learning approaches. Eventually, we found a path for abstracting large, complex datasets into structures that AI agents could actually reason over and work with.
As we built, we started working across other domains and saw the same patterns repeat.
They showed up in climate data through our work with NCICS, the North Carolina Institute for Climate Studies, analyzing petabytes of NOAA data. They showed up in subsurface data like 3D seismic cubes. They showed up in non-destructive testing, sensor data, and other large multimodal datasets across advanced industries.
That’s what Lium is built for.
We’re abstracting away the complexity of AI engineering and data engineering for these kinds of datasets, and building an agentic workspace purpose-built for complex data science.
For subject matter experts, Lium makes it possible to work with data and get insights without needing to be a software engineer or data engineer.
For data teams, Lium significantly extends what they can do and how much ground they can cover.
And because Lium is built as a collaborative workspace, teams can share workflows, build on each other’s work, and compound their institutional knowledge over time. We sometimes describe it as: if Cursor and Notion had a baby, it would be named Lium.
Ultimately, we’re building Lium to help organizations turn complex data into discoveries and, hopefully, real positive impact in the world.
We’ve poured our hearts into this, and we’d genuinely love your feedback.
Please give Lium a try and let us know what you think.
Explore more artificial intelligence tools on IndieHunt.
