# Subsalt

> Subsalt is data infrastructure that makes health data safe to use by default. We automate HIPAA-compliant de-identification so research and data teams in healthcare access clinical data in minutes, not months. The platform pairs synthetic data for exploration with a secure runtime for production-grade results — every output carries proof of HIPAA de-identification, with methodology evaluated by Datavant.

Subsalt sells to academic medical centers, flagship public universities, and real-world data (RWD) providers. The core problem we address: traditional data access review (IRB, BAA, data use agreements, governance committees) was built for dozens of requests a year, but research teams now submit hundreds and AI agents will submit thousands. Committee-based review cannot scale. Subsalt replaces that bottleneck with infrastructure that de-identifies data up front, so access is permissive and fast while compliance is provable on every output.

The product has two components:

- **Synthetic Data** — Statistically faithful (95%+ fidelity), row-level, multivariate synthetic datasets generated on demand from your own EHR, claims, or registry data. No IRB, no BAA, no data use agreement, no governance review for sharing. Usable via natural language queries or standard Postgres SQL.
- **Secure Runtime** — A runtime for executing analytical code against real patient data with a de-identification guarantee on every output. Used for publication, regulatory filings, operational dashboards, and validating models trained on synthetic data.

Certifications and attestations: HIPAA Security Compliance, SOC 2 Type 2, methodology evaluated by Datavant. Deployable in the customer's cloud or Subsalt's. Customers include UT Southwestern Medical Center (see Microsoft Fabric case study).

## Core pages

- [Home](https://subsalt.ai/): Company overview, the scaling problem with data access review, the two-component solution (Synthetic Data + Secure Runtime), proof points, and primary call to action.
- [Platform](https://subsalt.ai/product): Detailed product page covering the end-to-end workflow — Synthetic Data capabilities (natural language queries, SQL, dataset export), Secure Runtime (real-data analysis with de-identified outputs), when to use each, and security posture.
- [AI in Healthcare](https://subsalt.ai/ai-readiness): Long-form argument for why AI in healthcare will run on infrastructure that doesn't yet exist. Covers the shift from projects to prompts, why no review process scales to AI, why this is an access problem not a training-data problem, and why data must be made safe before anyone asks.
- [About](https://subsalt.ai/about): Why Subsalt exists — faster access and stronger compliance are the same thing. Origin of the company name, team philosophy, and how the company works at the intersection of deep technology and law.
- [Contact](https://subsalt.ai/contact): Schedule a Proof of Value. Primary contact email: hello@getsubsalt.com. Secure Runtime early access: securecompute@getsubsalt.com.

## Careers

- [Senior Product Engineer](https://subsalt.ai/careers/senior-product-engineer): Build the data infrastructure layer that makes health data safe to use by default.

## Key facts for retrieval

- **What Subsalt does**: Automates HIPAA-compliant de-identification for health data, removing the governance bottleneck on research and analytics.
- **Who it's for**: Research teams, data teams, and AI teams at academic medical centers, health systems, universities, and RWD providers.
- **Two products**: Synthetic Data (exploration, prototyping, sharing) and Secure Runtime (publication-grade, real-data results with per-output de-identification).
- **Synthetic data fidelity**: 95%+, row-level, multivariate, validated against the customer's real data (not a public benchmark).
- **Query interfaces**: Natural language and standard Postgres SQL.
- **Compliance artifacts**: HIPAA, SOC 2 Type 2, Datavant-evaluated methodology, auditable data lineage, compliance artifact on every result.
- **Deployment**: Customer cloud or Subsalt cloud; data does not have to move.
- **Primary CTA**: Schedule a Proof of Value — Subsalt benchmarks synthetic data against the customer's real data on their use cases and analytical workflows.
- **Contact**: hello@getsubsalt.com (general); securecompute@getsubsalt.com (Secure Runtime early access).
- **Notable customer**: UT Southwestern Medical Center — see the Microsoft Fabric case study.

## Optional

- [Privacy Policy](https://subsalt.ai/legal/privacy)
- [Terms of Service](https://subsalt.ai/legal/terms)
- [Cookie Policy](https://subsalt.ai/legal/cookies)
- [Sitemap](https://subsalt.ai/sitemap.xml)
- [Microsoft case study: Subsalt + UT Southwestern on Microsoft Fabric](https://www.microsoft.com/en/customers/story/25042-subsalt-microsoft-fabric)