Synthefy

foundation-model

enterprise

Published

June 12, 2026

Overview

Synthefy, Inc. is an Austin-based startup (founded 2023) building tabular foundation models for structured-data prediction, positioning its technology as a drop-in replacement for traditional tools like XGBoost. The company aims to deliver production-ready predictions through a single API call, eliminating feature engineering, model selection, and hyperparameter tuning.

Key Product

Nori (Synthefy-Nori-V1) is a tabular foundation model for regression via in-context learning. Pretrained from scratch on millions of synthetic datasets generated from causal graphs and regression trees, it predicts a full set of quantiles to return calibrated prediction intervals. At ~6M parameters, it runs on a single GPU with ≥8 GB VRAM and is released open-weights under Apache-2.0.

Team

Founded in 2023 by Sandeep Chinchali, Shawn Jain, Raimi Shah, and Somi Agarwal.

Outlook

Synthefy targets enterprise structured-data prediction with customers including Samsung, NetApp, and Deutsche Telekom, and has secured a US Army Phase 2 SBIR award and a NeurIPS 2025 paper acceptance. As a seed-stage company, broad adoption against entrenched gradient-boosting tools and other tabular foundation models will be the key test.