Analysts and researchers are constantly seeking ways to unlock deeper insights from their datasets. But what if you could do this while simultaneously upholding the highest standards of data privacy and security? At Duality, we’re excited to introduce our new solution: Duality Assistant Agent, an innovative framework designed to transform how you conduct a secured collaborative research.
Imagine having a dedicated research assistant that truly understands your data, your unique workloads, and your research goals while ensuring that sensitive data is not exposed to any party. That’s exactly what Duality Agents are. These intelligent, agent-based assistants sit seamlessly on top of the Duality platform, learning the nuances of your collaboration environment and using privacy preserving technologies to protect your data processing and outcome.
They’re not just tools; they’re intelligent partners, proactively guiding you through complex analyses and helping you extract meaningful insights. By leveraging Duality Agents, you can make your research journey more intuitive, efficient, and impactful, freeing you to focus on the breakthroughs that matter most.
Until now, users interacted with Duality through a traditional interface for running sessions. As a researcher, you would select your project, navigate to the session page, and manually configure parameters such as workload name, workload inputs, and filters.
With the introduction of the Duality Assistant Agent, users can now engage with a chat-like interface using plain English — marking a significant shift in user experience. Beyond the convenience of natural language, this new approach empowers users to unlock insights that were previously difficult to achieve.
Through the chat interface, users can ask questions and explore their collaboration data and workloads as naturally as interacting with ChatGPT — all within the boundaries of Duality’s privacy and governance controls, ensuring that only authorized data is accessed.
In addition to conversational interactions, the Assistant also supports deep research: users can define a research question or problem, and the system performs a comprehensive analysis across the relevant data and workloads, returning a clear, summarized report.
As a user, you can configure and enrich your agent by uploading any file and data (e.g. research document) relevant to your analysis. The agent will learn this data to fine tune your research plan and analysis needs..
Examples:
Here’s an example of a typical research task. Behind the scenes, the agent securely runs workloads based on the user’s prompts.
Start from scratch:
I’m a pediatric cancer researcher interested in studying survival rate. What are the datasets and all the analysis available for me?
Run a specific analysis per the agent suggestion (e.g. survival rate)
The result highlights a specific cancer type with higher mortality. As a next step, we ask the agent to recommend further analysis
Duality Agents are engineered to empower you with capabilities that were once out of reach. Here’s how they’ll revolutionize your workflow:
At the heart of secure, collaborative research lies the need to harness AI capabilities while protecting sensitive data. The Duality Assistant Agents is purpose-built to meet this need – enabling AI-driven research support across organizational boundaries without compromising data privacy or security.
The system consists of three tightly integrated components:
1. Research Assistant AI Agent
This is the AI interface researchers engage with directly. Each agent operates under a well-defined role and task within the research project – for example, planning required analytics to address a research question, executing analytics, analyzing model outputs, or synthesizing results summaries. Crucially, the AI agent does not have unrestricted access to sensitive data. Instead, it issues requests within a secure framework governed by strict privacy and entitlements. The agent acts as a productive assistant while respecting all data protection requirements.
2. Duality MCP Server (Model Context Protocol)
The MCP server plays a critical intermediary role. It manages the tools, resources, and interaction context available to the AI agent, all in accordance with the current user’s permissions and project state. Through the Model Context Protocol, it provides structured, policy-aware interfaces that allow the AI agent to request computations or insights without ever accessing raw data. It ensures that all AI interactions are grounded in the specific user research context and are compliant with collaboration rules.
3. Duality Platform
This is the core infrastructure that manages the overall collaboration between the participants. It controls data entitlements, governs workloads, and ensures that each research activity follows the proper governance and security model. The platform also orchestrates the use of advanced privacy-enhancing technologies, including homomorphic encryption, federated analytics and learning, differential privacy, and secure enclaves, to protect sensitive data across all participating organizations. The platform facilitates the secure computation of insights either encrypted or where data resides, ensuring only privacy-preserving results are shared.
Together, these components form a secure, scalable system that empowers researchers to use AI assistants effectively while maintaining strict privacy guarantees. With the Duality Assistant Agent system, institutions can collaborate on complex research questions, even when the underlying data must remain siloed and confidential.
At Duality, privacy isn’t an afterthought; it’s fundamental to everything we do. Duality Agents are powered by the most advanced privacy-preserving technologies available, ensuring that your sensitive data remains protected throughout the entire analysis process.
When an agent is running, it operates within the Duality platform, which leverages advanced privacy-preserving technologies — including fully homomorphic encryption (FHE), trusted execution environments (TEE), federated learning, and differential privacy — to ensure the security and confidentiality of both data and computations.
Under the hood, Duality ensures that only authorized data and workloads can be analyzed. Crucially, your data remains encrypted during computation, ensuring there is absolutely no data leakage. This unwavering commitment to privacy means your collaborations are always secure, even when leveraging the most powerful AI capabilities for deep research.
Use of LLMs
Duality supports multiple modes of integrating LLMs within its platform.
In most cases, LLMs are used strictly to help generate the necessary API calls for executing sessions within Duality — no sensitive data is shared with the LLM in these scenarios.
However, when a user submits prompts that explicitly require LLM-based analysis of results (such as in deep research workflows), the system offers two options:
Duality Agents represent a significant leap forward in secure, collaborative data analysis. They combine cutting edge AI with Duality’s robust privacy-preserving technology to offer an assistant that is both powerful and trustworthy.
Ready to see Duality Agents in action and transform your research? Reach out to us for a personalized demo!
Duality Assistant Agents represent a step-change in how teams analyze sensitive data. They make advanced, privacy-preserving research accessible, intuitive, and scalable.
