UNLOCKING BIOMED INNOVATION

Driving better research.

Rexis streamlines the research process by abstracting away the complex data acquisition and vetting process, and then providing all of the resources that you need to train and fine-tune any biomedical model on this data.

Get Started

Rexis boasts thousands of datasets

Built for labs, clinics, companies, and institutions.

The END-TO-END SOLUTION

Encrypted data, fine-tuning, and AI inference in one place.

Rexis is the first platform to vertically integrate into biomedical data pipelines right through to training and inference execution, without ever revealing underlying data.

Cryptographically secure encryption

Quantum-safe encryption

Fast, performant training compute

Private data lockers

How EE-protected model training and fine-tuning works on Rexis.

Pick one or more datasets

Rather than transferring raw data, users issue compute-to-data queries that execute securely on encrypted datasets, governed by privacy-preserving smart contracts.

Pick a biomedical AI model

Rexis supports the most biomedical AI models in the industry across modalities like Systems Biology, Metabolic Modeling, Spatial Transcriptomics, and Structural Bioinformatics.

Start the fine-tuning process

Each institution securely computes local gradient updates on its own encrypted EHRs, and a decentralized aggregator combines these updates to produce a refined model.

Automate model benchmarking

No plaintext data ever leaves institutional control, yet cryptographic commitments and lightweight proofs attest to the correctness of each inference.

Run AI inference on new model

All proprietary model details including model weights remain encrypted by Equivariant Encryption, while underlying data prompts are passed to the model in an encrypted state.

The ultimate

data

ecosystem

Data is uploaded, containerized and encrypted

Rexis supports all data types and biomedical modalities for upload and usage.

Data remains encrypted during model training

Rexis maintains end-to-end encryption on all underlying datasets used during fine-tuning.

The resulting model is encrypted and works on encrypted data

Rexis then facilitates AI inference of the resulting encrypted model on encrypted data.

Validating AI models in biomedicine demands rigorous testing on diverse, real-world data. The Rexis verification process enables institutions to evaluate newly developed algorithms (e.g., cancer diagnostics) on encrypted patient records from multiple independent sources for robust performance metrics across varied populations and clinical settings.

RUN YOUR MODEL

Rexis supports EE-secure AI inference for your trained model directly in the platform.

All AI inference is protected by a breakthrough privacy technology, EE.

Read about EE

Read case study

Equivariant Encryption (EE) is fully private like Homomorphic Encryption (HE), but with 1000x less latency.

Read EE vs HE

Read case study

EE privacy-preserving technology is compatible with all major biomedical model modalities.

See EE Coverage

Try EE for yourself through the open repository.

Run the demo code

Read case study

GET STARTED WITH REXIS

Unleash your scientific research with Rexis.

Get Started

Buy Template

Book an intro call