Technology
Tabular foundation models
Pre-trained deep learning models that deliver instant, state-of-the-art predictions on structured data without task-specific training.
For two decades, gradient-boosted decision trees like XGBoost ruled tabular machine learning, requiring extensive hyperparameter tuning and custom pipelines for every new dataset. Tabular Foundation Models (TFMs) like TabPFN disrupt this paradigm by applying transformer architectures to structured data. Trained on millions of synthetic datasets, these models use in-context learning to perform classification and regression in a single forward pass. By eliminating the need for manual feature engineering or training loops, TFMs deliver highly accurate predictions on small-to-medium datasets in seconds, fundamentally shifting how teams deploy machine learning for spreadsheets.
What builders pair with Tabular foundation models
Projects using both technologies. Select a pairing to see a project.
Pairing: Claude Haiku
Teaching an Agent to Behave Like a Data Scientist: A Statistical Harness for Tabular ML
Pairing: Claude Sonnet 5
Teaching an Agent to Behave Like a Data Scientist: A Statistical Harness for Tabular ML
Pairing: LangGraph
Teaching an Agent to Behave Like a Data Scientist: A Statistical Harness for Tabular ML
Pairing: LLM
Teaching an Agent to Behave Like a Data Scientist: A Statistical Harness for Tabular ML
Recent Talks & Demos
Showing 1-1 of 1