Wing Venture Capital leads investment in a foundation-model company bringing intelligence to the structured data that runs the physical and digital economy
SAN FRANCISCO, August 18, 2026 — Synthefy today unveiled its foundation-model platform for structured data. The company’s Structured Data Foundation Models (SDFMs) bring foundation-model intelligence to the numbers, tables, transactions, and signals that drive the physical and digital economy without building and training a separate model for every problem.
Synthefy also announced a $6.5 million seed round led by Wing Venture Capital, with participation from Haystack, Samsung Next, Canonical, and Lightscape. The round also includes investments from Srinivas Narayanan, former CTO of Applied AI at OpenAI; Aparna Chennapragada, Chief Product Officer of Experiences and Devices at Microsoft; and Manohar Paluri, Vice President of AI at Meta Superintelligence Labs.
“While much is focused on artificial general intelligence that replicates human capabilities through language, we’re building machine intelligence for problems humans cannot solve at scale,” said Somi Agarwal, co-founder of Synthefy. “Our models are designed specifically for structured numerical data, bringing foundation-model intelligence to the numbers, tables, transactions, and the signals that drive the global economy.”
Bringing foundation models to the world’s structured data
LLMs learned reusable patterns by pretraining on vast collections of text, and Synthefy applies the same approach to structured numerical data. Synthefy’s Structured Data Foundational Models process structured data tokens while preserving relationships within tables and time series. A structured data token can represent a sensor reading, a financial transaction, a stock tick, a housing price, a meter pulse, or a biometric signal.
Unlike traditional machine learning, where every dataset and use case demands its own purpose-built model, Synthefy’s Structured Data Foundational Models generalize across numerical data itself. A pretrained model can forecast demand across thousands of products, optimize prices in real time, detect financial fraud, anticipate infrastructure failures, and improve data center efficiency. The same foundation model can extend to any problem that can be expressed as structured data.
Designed for enterprise deployment and control
Recent advances in architecture, data, and computing have made numerical pretraining practical at scale. Synthefy turns that capability into a reusable foundation-model layer for forecasting, pricing, fraud detection, risk analysis, and infrastructure monitoring.
Enterprises and developers can generate production-grade predictions through an open model, a managed API, an enterprise platform, or a deployment inside their own environment. The platform supports the data environments enterprises already use, including AWS, Databricks, Snowflake, and Google Cloud.
Synthefy is already working with customers and partners across retail, financial services, telecommunications, infrastructure, healthcare, and defense. Its models are being used for demand forecasting, pricing optimization, risk detection, and infrastructure failure prediction.
Nori brings foundation-model intelligence to tables
Synthefy’s models are smaller and purpose-built for numerical prediction and require less compute and energy than large, general-purpose language models.
Nori is an open-source structured data foundation model. Nori helps enterprises solve high-value problems, including pricing optimization, fraud detection, risk analysis, and demand forecasting.
- Nori ranks first across 130 public (regression) benchmarks.
- Available under an Apache 2.0 license, almost 600K Nori models have been downloaded, and 5,000 Python installations have occurred in its first weeks since launch last month.
“The rise of foundation models for structured data represents the next major expansion of artificial intelligence,” said Gaurav Garg, founding partner at Wing Venture Capital. “Synthefy is building a model platform that can address some of the largest and most valuable datasets in the world. Its combination of technical performance, efficient architecture, open models, and early enterprise adoption positions the company to define this emerging category.”
Synthefy will use the new funding to expand its research and engineering teams, develop the next generation of Nori, and establish partnerships in industries where better numerical prediction can generate significant economic and operational value.
“Our vision is a single model layer for the world’s structured data, one foundation that any team can point at any numerical problem,” said Assistant Professor Sandeep Chinchali, University of Texas at Austin, and co-founder of Synthefy. “Foundation models for structured data will become as important as foundation models for language. They will be less visible and less conversational, but they will quietly improve the decisions that move value through the real world.”
About Synthefy
Synthefy is building Structured Data Foundation Models, foundation models designed specifically for the numerical data that runs the world. Its open, enterprise-ready models help organizations forecast demand, optimize pricing, detect risk, predict failures, and simulate outcomes without building and training a separate model for every problem.
Founded in 2023, Synthefy is backed by Wing Venture Capital, Haystack, Samsung Next, Canonical, Lightscape, and leading artificial intelligence executives. The team includes former engineers and scientists from Meta, Nvidia, Adobe, Uber, and leading academic institutions, with research published at top conferences.
Learn more at www.synthefy.com.
Contact: Michael Celiceo, [email protected]
SOURCE Synthefy
