Kumo’s cover photo
Kumo

Kumo

Software Development

Mountain View, California 16,669 followers

Build AI models for relational data — without feature engineering.

About us

Kumo is an AI model company that provides a platform for anyone to train and run state-of-the-art AI models on their relational data. Models built on Kumo already power the recommendations, fraud detection, risk scoring, entity resolution, RAG, and other mission-critical components inside applications used by hundreds of millions of people globally. The company is led by pioneers in Relational Deep Learning and Graph Transformers with the mission of helping AI teams make their most important data also their most useful data.

Website
http://kumo.ai
Industry
Software Development
Company size
51-200 employees
Headquarters
Mountain View, California
Type
Privately Held
Founded
2021
Specialties
Artificial Intelligence and Machine Learning

Locations

  • Primary

    357 Castro St

    Suite 200

    Mountain View, California 94041, US

    Get directions

Employees at Kumo

Updates

  • Kumo reposted this

    The best teams work hard and play harder. Last week our team gathered in Berlin for a meetup — and this little behind-the-scenes clip captures exactly why I love this group. At Kumo, we think a lot about networks. In a graph, a single node can only do so much. But connect those nodes together and something remarkable happens — the whole becomes far greater than the sum of its parts. That's exactly what "better together" means to us. Not just a value we put on a slide, but something you feel when you're in a room with this team. The energy, the late nights, the laughter, the debates about the right way to solve a hard problem — this is what building something meaningful actually looks like. Proud of every single node in this network. 🙌 Ramona Bendias and Keti Jovanovska crushed it on organizing this. Special thanks to our speakers Matthias Fey, Jingang Qu! #Kumo #BetterTogether #TeamCulture #Berlin #AI #GraphAI #StartupLife

  • View organization page for Kumo

    16,669 followers

    We are pleased to share a video featuring interviews with attendees from our recent #Berlin meetup, where they discuss their perspectives on foundation models and how they see the future evolving. Thank you to everyone who joined us and contributed to the conversation. Your insights and engagement made the event truly valuable. We look forward to welcoming you to our next meetup. Stay tuned for upcoming announcements. Try Kumo today: https://lnkd.in/dkgPS4y3

  • View organization page for Kumo

    16,669 followers

    What if AI did not just analyze the past but predicted what happens next? Our CEO and Co-founder Vanja Josifovski joined Andy Walsh on Startups Decoded to discuss the next phase of #AI, #prediction. From using enterprise data to forecast outcomes to automating real business decisions, the conversation explores how companies can move from insights to action. In this episode: • Why predictive intelligence is the next phase of AI • How enterprises can use their own data to forecast outcomes • Where trust breaks down in AI adoption • What happens when machines outperform humans in operational decisions Listen here: Apple: https://lnkd.in/dp2T9waF Spotify: https://lnkd.in/dweUniAc YouTube: https://lnkd.in/dD5Yxhh8

  • View organization page for Kumo

    16,669 followers

    Huge thank you to everyone who joined us yesterday at the Kumo #Berlin Tech Meetup! 🙌 We had an incredible time exploring how modern AI is evolving beyond traditional approaches. The discussion highlighted a new generation of foundation models for tabular and relational data, unlocking powerful capabilities in prediction, recommendation, and large-scale analytics. Together, we explored: • How relational foundation models transform complex datasets into intelligent systems • How these models integrate with cloud data warehouses and modern data infrastructure A special thank you to our amazing speakers: • Jingang Qu, Postdoctoral Researcher at Inria and author of TabICLv2 • Matthias Fey, Founding Engineer at Kumo and creator of PyTorch Geometric • Miha Pelko, ML Infrastructure Engineer at Kumo And of course, a big shoutout to our team who made this event possible: Ramona Bendias, Federico Lopez, Keti Jovanovska👏 We will keep you posted on what is coming next!

    • No alternative text description for this image
    • No alternative text description for this image
    • No alternative text description for this image
    • No alternative text description for this image
    • No alternative text description for this image
      +1
  • View organization page for Kumo

    16,669 followers

    Huge thank you to everyone who joined us yesterday at the Kumo Berlin Tech Meetup! 🙌 We had an incredible time exploring how modern AI is evolving beyond traditional approaches. The discussion highlighted a new generation of foundation models for tabular and relational data, unlocking powerful capabilities in prediction, recommendation, and large-scale analytics. Together, we explored: • How relational foundation models transform complex datasets into intelligent systems • How these models integrate with cloud data warehouses and modern data infrastructure A special thank you to our amazing speakers: • Jingang Qu, Postdoctoral Researcher at Inria and author of TabICLv2 • Matthias Fey, Founding Engineer at Kumo and creator of PyTorch Geometric • Miha Pelko, ML Infrastructure Engineer at Kumo And of course, a big shoutout to our team who made this event possible👏 We will keep you posted on what is coming next!

  • Kumo reposted this

    Reminder: Tomorrow we’re hosting a Kumo meetup in Berlin. 🚀 We’ll talk about relational foundation models, graph ML, and what it actually takes to run AI on large-scale structured data in production. I’m especially happy to have Jingang Qu, Matthias Fey & Miha Pelko speaking. We pulled this together in about three weeks, and I’m very excited to see such strong interest from the Berlin technical community. If you’re in Berlin and work on ML, data, or engineering systems, join us: https://luma.com/6solbl9x #BerlinAI #GraphML #DataEngineering

    • No alternative text description for this image
  • View organization page for Kumo

    16,669 followers

    Great perspective from Hema Raghavan. These issues with PyTorch and XGBoost are exactly what we see when agents operate at the wrong level of abstraction. Kumo is designed to give agents the right foundation for relational data, so they can produce reliable results with far less complexity.

    Coding agents are only as good as the abstractions they work with. We put this to the test using Kumo's Research Agent to build an account-level fraud model — and the results were eye-opening. When coding agents work with low-level frameworks, mistakes are almost inevitable. PyTorch errors: double class imbalance, loss averaged by data size instead of sample size XGBoost errors: label granularity issues, label leakage These aren't careless mistakes — they're a symptom of working with the wrong abstractions. Low-level frameworks weren't designed for the way agents reason about problems. With Kumo's Research Agent, the same task produced zero errors — and required just 55 lines of code instead of 1,014. The resulting model was also 10% more accurate than an expert-built XGBoost model in production. The lesson? The abstraction layer matters as much as the agent itself. Kumo gives coding agents the right foundation to reason about relational data — so they stop making the kinds of mistakes that quietly undermine your models. We're seeing this play out across fraud detection, recommendations, risk scoring, and more. The teams winning with AI aren't just picking the best agent — they're pairing it with the right platform. #AI #MachineLearning #CodingAgents #MLOps #GraphML #Kumo

    • No alternative text description for this image
  • View organization page for Kumo

    16,669 followers

    Join us at the Kumo #Berlin Tech Meetup to explore the future of relational foundation models, systems, and real world applications! What to expect: • Technical deep dives into relational foundation models • Practical insights into real world applications and system design • Networking with engineers, researchers, and AI professionals We are #hiring, and you will have the chance to meet the team and learn more about open roles. Learn more and register: https://luma.com/6solbl9x

    • No alternative text description for this image

Similar pages

Browse jobs

Funding

Kumo 2 total rounds

Last Round

Series B

US$ 18.0M

See more info on crunchbase