Korea's Nums AI adds $1.49M pre-seed as it chases the tabular-data AI race
What's the deal? Nums AIDealroom has a profile for this one. Try Dealroom →, a South Korean startup building foundation models for structured business data, has closed a second pre-seed tranche of US$1.68M (roughly $1.49 million). Mirae Asset Venture InvestmentDealroom has a profile for this one. Try Dealroom → and Korea Investment PartnersDealroom has a profile for this one. Try Dealroom → each contributed US$838.4K, the company said on November 4.
What's the endgame? Nums AI develops a tabular foundation model (TFM) that runs predictions on rows-and-columns data — sales records, transactions, customer details. Its model, Causilo, targets questions like churn risk and demand forecasting, aiming to turn stored numbers into business decisions.
Why now? The round follows a US$3.35M first raise in July from Stonebridge VenturesDealroom has a profile for this one. Try Dealroom →, SBVA, KT InvestmentDealroom has a profile for this one. Try Dealroom →, and Base VenturesDealroom has a profile for this one. Try Dealroom →. That lifts total funding to US$5.03M. The company also won a spot in TIPSDealroom has a profile for this one. Try Dealroom →, the government's tech startup support program, securing a further US$670.7K in research funding.
What's the money for? The capital will fund model development and hiring. "In the increasingly competitive field of structured-data AI, speed matters more than anything," said chief executive officer Yoo Jae-min (translated from Korean), adding that the new funding would accelerate development.
Founded in February by Seoul National University professor Yoo Jae-min, Nums AI drew early attention for raising a large pre-seed on team credentials alone. Causilo ranked first across overall, classification, and regression tests on the TabArena benchmark at its September debut, and was cited alongside Nvidia's Kumo-Tabular and Prior Labs'Dealroom has a profile for this one. Try Dealroom → TabPFN-3.5 during Jensen Huang's keynote at NVIDIA GTC Berlin 2026.
What could go wrong? The raise and TIPS selection extend Nums AI's runway, but commercialisation remains unproven. The company is running proofs-of-concept with clients including InBodyDealroom has a profile for this one. Try Dealroom →, yet customer traction and market validation are still to come.
The signal: As AI investment crowds around language models, investors are backing specialist models aimed at the tables and figures that make up much of enterprise data. Nums AI's follow-on round is a bet that structured data is the next foundation-model battleground.
Read more: wowtale.net
Image credit: Nums AI
