The Devron Platform
Instead of bringing mountains of data to the model, Devron brings the model to the data—training algorithms where the data resides. As a result, organizations can accelerate their data science efforts, unlock trapped insights, and realize better business outcomes.
Federated Data Science
Devron is a federated data science and machine platform that enables teams to build and train AI models on distributed, heterogeneous, and private data where it resides.
Deploy Across Clouds
Devron is an easily deployable enterprise platform. Its flexible architecture allows users to implement across multiple accounts within the same cloud provider or across multi-cloud providers within your company or between different companies.
We offer flexible deployment models, including installing in our customer’s cloud environment, a Devron-hosted environment, or a hybrid of both, as well as on-premise. It’s also quick to deploy—fully operational in only a few hours.
Devron-Hosted
Use a Devron-hosted cloud environment dedicated to your company.
Major Cloud Providers
Deploy inside VPCs on AWS, GCP, and Azure.
On-
Premise
Deploy on-premise using your existing infrastructure.
How It Works
The Devron platform consists of two main components: a Control Center and one or more Satellites.
The Control Center is the client interface where data scientists can explore datasets, develop models, and run experiments.
Each Satellite is proximate to the data source and connects to a dataset in a read-only, privacy-preserving manner, training models developed and distributed from the Control Center.
As each satellite completes local model training, it sends model weights and artifacts back to the Control Center for aggregation. Raw data never transfers from the Satellites to the Control Center—only model metadata.
Client Interfaces
Devron offers two different user interfaces, depending on the level of data science sophistication, and a separate data owner interface for dataset provisioning and privacy control.
Devron
Python SDK
Data Scientists & Machine Learning Engineers
Use a Notebook interface via our Python SDK for greater customization and control.
Devron
Flow™
Business Analysts & Citizen Data Scientists
Devron Flow™, our low-code interface, provides a drag-and-drop model-building experience for non-technical users.
Devron
Privacy Portal
Data Owners & Providers
Our Devron Privacy Portal allows data owners to specify custom privacy controls that the Devron system will enforce when engaging with that specific dataset.
Flexible APIs to enable a federated data science experience
With Devron’s SDK, developers can integrate their own proprietary enterprise warehouses and homegrown applications to build powerful federated machine learning applications.
Devron Benefits
Devron enables data science teams to realize enhanced outcomes, improved data science agility, and reduced overhead—all without moving or exposing data.
Realize Better Business Outcomes
Unlock access to more data, including datasets containing PII and PHI.
Gain better insight by boosting model accuracy, generalizability, and reliability.
Gain Data Science Agility
Bypass lengthy data movement and approvals processes.
Test data value hypotheses in a matter of hours instead of days or weeks.
Reduce Overhead & Risks
Reduce overhead of data movement, duplicate infrastructure, & ETL pipelines.
Lower cybersecurity risk and privacy leakage concerns.
Preserve data privacy, staying compliant and secure—only sharing model learnings, not the data.
Reducing Friction Between Data Owners & Consumers to Create More Value from Data
Devron has established a new AI and data science paradigm to help companies better realize the full potential of their data while setting a new standard for data privacy.
By putting data owners and producers in control of their data’s privacy, Devron allows business units, regional teams, customers, and partners to share data securely and privately.
At the same time, Devron allows data consumers to easily explore, transform, analyze, and model valuable datasets without ever seeing the underlying source information.
Built with Privacy & Security at the Core
Forged in the most data-sensitive environments of the Defense and Intelligence communities, Devron is designed to meet the strictest regulatory and compliance requirements, including GDPR and CCPA. Devron is also HIPAA and SOC2 Type 2 certified.
Additionally, the platform leverages a hybrid combination of privacy-enhancing technologies to ensure strong data privacy, including partially homomorphic encryption, secure multi-party computation, synthetic data, differential privacy, and k-anonymity style measures.
The combination of these capabilities allows users to unlock previously-inaccessible datasets, shorten the time to access data, accelerate experimentation and time to insight, reduce the risk and overhead of moving data, and produce analytics of greater accuracy and effectiveness.
API-Based Privacy
Return Results, Not Raw Data
Synthetic Data
Allows for exploratory data analysis without exposing raw data.
Multi-Dataset Alignment
Privately map relevant PIIs between datasets.
Federated Querying
Run queries against multiple siloed data sources without unifying the data.
Vertical Training
Train models over vertically split (heterogeneous schema) data sources.
Privacy Technologies
Flexible for Multi-Layer Protection
Secure Multi-Party Computation
Distributes a computation across multiple parties where no individual party can see the other parties’ data.
Partially Homomorphic Encryption
Aligns personally identifiable information across data sources.
K-Anonymity
Prevents privacy leakages from computation on small amounts of data.
Differential Privacy
Hides metadata information and statistics critical for effective data analytics.
Cryptographic Hash Functions
Translates data of various lengths into a fixed size numerical string—the hash.
Fits Your Data Stack
Multi-Cloud
Devron is cloud-agnostic and can be seamlessly run within your existing cloud environments, including AWS, Microsoft Azure, and Google Cloud.
Integrates with Common Services
Devron integrates with common data lakes, warehouses, and data processors, including Snowflake and Databricks, and is compatible with common cloud services, such as Azure Synapse, BigQuery, Amazon RDS, and Amazon Redshift.
Enhanced & Scalable
Devron leverages Enterprise Cloud ML solutions to govern and manage jobs across multiple clouds for an enhanced and scalable federated ML experience.
Developer Ergonomics
Our Python SDK provides developer ergonomics to integrate with all your favorite data science tools.
Getting Started with Devron
Devron’s engagement model consists of four simple steps:
1. Identify Use Case & Data Sets
Work with Devron forward-deployed engineers to refine use cases and help identify valuable data.
2. Deploy Devron Software
Deploy software (or leverage automated cloud deployment) and connect to relevant data stores.
3. Develop Analytics
Begin leveraging distributed datasets for model development, experimentation, and training.
4. Gain
Insights
Serve the model, turn info into intel, and present findings using tools, like Qlik Tableau, & PowerBI.