Fundamental

The NCAA puts the odds of a perfect March Madness bracket at 1 in 120 billion. Every year it’s one of the most-attempted prediction problems in America, and it's a tabular one. So we proposed a session at SXSW to run at it directly. Man vs. Machine Madness: Can AI Beat Your Bracket? Fundamental's Large Tabular Model, NEXUS, against world-class athletes. Same field, same blank bracket, filled live on stage. NEXUS knows decades of tournament games. The athletes have lived the pressure that decides them. Then March settles it. Whether it happens is up to the community. SXSW PanelPicker voting is open now through August 23. --- 👉 Create an account: https://lnkd.in/gsBSE2k 👉 Confirm your email 👉 Open the session page and click the ❤️ next to vote: https://lnkd.in/gnvm8vJm

This weekend our Co-Founder & CEO Jeremy Fraenkel joined AI leaders at Stanford University for a panel on The Future of AI Beyond the Chatbot Era Alongside Igor Babuschkin (Co-founder xAI), Ricardo Baeza-Yates (Search Chief Scientist, You.com), and Lu Zhang (Founder Fusion Fund), the discussion explored where AI goes next: from specialized systems and autonomous agents to trust, ownership, and what comes after today's generation of chatbots

Diffusion models didn't come out of nowhere. They're a new lens on ideas researchers have been circling for years. In the latest episode of First Principles Sander Dieleman (a 10+ year veteran of Google DeepMind) talks about why so many “separate” ML techniques turn out to be the same idea in disguise, and why human intuition falls apart once you’re reasoning about probability in high-dimensional space. The full interview is linked in the comments.

One of our ML Researchers Kevin Scaman recently presented at the International Conference on Machine Learning (ICML). Kevin shared some findings from his upcoming research on ‘What LLMs Learn (and Don't) from Tables’, highlighting how LLMs can perform with some tabular tasks, but they rely heavily on memorization and their predictive performance degrades as tabular complexity increases. Research like this helps advance our understanding of where today's models excel, where they fall short, and why enterprise prediction requires approaches purpose-built for structured data. Thank you to [ICML] Int'l Conference on Machine Learning for bringing together the research community.

Last night, Fundamental hosted a great event with AICamp in San Francisco. The conversation explored where LLMs fall short on structured data, how Large Tabular Models differ, and what it takes to deploy AI in production. A special thank you to our guest speakers, Remy Thellier (AI/ML Partnerships, Snowflake) and Purna Sanyal (Generative AI, Amazon Web Services (AWS)), for joining our CEO Jeremy Fraenkel, Head of Applied AI Alexandre Gerbeaux, and Chief of Staff Tess Munsie, for a thoughtful discussion on the future of AI for structured data.

Fundamental is growing in Singapore. Singapore is one of the world's leading hubs for enterprise AI, where energy, logistics, commodities and financial services firms are moving fast on AI that delivers measurable business outcomes. To lead that work, Colin Tan Thong Tee has joined as General Manager, APAC, as we continue to build out our Singapore team to help customers put Large Tabular Models into production across some of Asia-Pacific's most data-intensive industries. Excited for what's ahead as we expand in the region. Alexandre Gerbeaux Tommy Yong Gabriel Suissa

This week, our Marketing Team descended on Fundamental HQ to strategize, plan for the year ahead and bond over dinners in San Francisco and wine tasting in Sonoma. From marketing campaigns to global events to activations in our backyard here in the Bay Area, we are working to educate the world about the power to predict with Fundamental’s Large Tabular Model, NEXUS. Want to join the team? Open roles are in the comments.

We're hosting a meetup with our friends at AICamp in San Francisco next week. Our CEO Jeremy Fraenkel and Head of Applied AI Alexandre Gerbeaux will join Remy Thellier (AI/ML Partnerships, Snowflake) and Purna Sanyal (Generative AI, Amazon Web Services (AWS)) to discuss where LLMs fall short on structured data, how LTMs differ, and what it takes to deploy them in production. If you're in the area, we hope to see you there. Registration link in the comments.

Before a ball was kicked, NEXUS predicted a Spain–Argentina final, with Spain lifting the trophy. The betting markets tipped France. Along the way, NEXUS also correctly identified all four semifinalists and finished with 84% accuracy on decisive matches, including a perfect quarterfinal round. Same public data. Different model. See how NEXUS tracked the tournament in the comments.

Mitsubishi UFJ Financial Group (MUFG). Japan's largest banking group and one of the largest financial institutions in the world today announced their strategic partnership with Fundamental. The deal will see MUFG deploy NEXUS across critical functions of their organization, unlocking huge value from their tabular datasets. https://lnkd.in/g4gus9hh