Lium (formerly Astromind) raises $5.5M seed to make complex scientific data LLM-readable
What's the deal? LiumDealroom has a profile for this one. Try Dealroom →, a Dallas-based startup formerly known as Astromind, has raised $5.5M in seed funding to launch an "agentic harness" that translates complex scientific datasets into formats large language models can actually understand. Investors include SJF VenturesDealroom has a profile for this one. Try Dealroom →, Wavemaker 360, Reach CapitalDealroom has a profile for this one. Try Dealroom →, and GC&H InvestmentsDealroom has a profile for this one. Try Dealroom →.
The core problem: LLMs handle text and code well but choke on satellite imagery, seismic surveys, and electromagnetic spectrum data. Scientists currently resort to tedious manual preparation to make their data AI-ready.
Lium's platform creates custom AI agents for each new data type it encounters. These agents restructure raw information so LLMs can ingest it and deliver consistent, reproducible results. The agents also improve over time, refining their formatting as more queries hit the dataset.
Why now? As AI models grow more powerful, the gap between what they can reason about and the data they can actually access is widening. Co-founder and chief executive Josh Knutson put it plainly: "The most important data across energy, science and infrastructure remains difficult to reason over."
Lium already has early traction in high-stakes settings. It has worked with astrophysicists to analyse sparse X-ray data — in one case examining exoplanet atmospheres to assist the search for extraterrestrial life. Closer to home, it processes terabytes of satellite and weather data for the North Carolina Institute for Climate Studies. Industrial power services company nexGENDealroom has a profile for this one. Try Dealroom → Inc. uses it to automate electromagnetic spectrum analysis for generator health reports.
What could go wrong? Hallucinations remain a persistent risk when LLMs interact with complex numerical data. Lium claims its structured approach reduces this problem, but proving reliability in scientific contexts demands a high bar of accuracy. The startup will also need to demonstrate that its agents generalise well across wildly different data types — astrophysics and soil science are very different domains.
The signal: SJF Ventures, which led the round, is an impact-focused fund with a track record in sustainability and climate-adjacent deals, underscoring that the near-term commercial pull for Lium likely sits in energy and environmental data rather than pure research. With Dealroom classifying Lium as "early growth" despite this being a seed round, the startup appears to have entered the market with paying customers already in place — a sign that demand for an AI-ready data translation layer is not theoretical.
Read more: SiliconAngle
