THE TECHNOLOGY BEHIND JARVIS
Use the Right Intelligence for the Right Task
Clinical trial automation requires more than sending a protocol to a large language model. Jarvis combines clinical knowledge, deterministic automation, purpose-built AI, specialized models, and selective LLM capabilities within an architecture designed for clinical data management.
Our Technology Philosophy
AI Where It Adds Value. Automation Where It Adds Control.
Different clinical data management tasks require different approaches. Some problems can be solved reliably through deterministic rules and structured processing. Others benefit from specialized AI models designed for a focused clinical task. Still others benefit from the language understanding and reasoning capabilities of larger models.
Jarvis is designed to use the appropriate approach rather than treating an LLM as the answer to every problem.
Deterministic Automation
Structured processing where consistency and control matter most.
Specialized AI (SLMs)
Focused models for a narrow, well-defined clinical task.
Selective LLM Capabilities
Language understanding and reasoning, used where it adds value.
Clinical Knowledge
Technology Informed by the Clinical Domain
Generic AI can recognize words. Clinical automation must understand what those words mean in the context of a study.
Jarvis incorporates knowledge about clinical study structure, Schedules of Activities, visits, assessments, data collection requirements, CRF design, Edit Check specification, standards, and organizational conventions. That knowledge informs both automation and AI-driven processes.
Intelligent SOA Processing
Understanding the Foundation of Study Data Collection
The Schedule of Activities provides a critical connection between the protocol and the study’s data collection design. Jarvis uses intelligent automation informed by clinical knowledge to detect and interpret SOA information, including study visits, timing, assessments, and related data collection requirements.
The objective is to transform information contained within the protocol into structured study information that can drive downstream EDC design.
AI, SLMs & LLMs
Fit-for-Purpose Models
Specialized Small Language Models offer the potential to perform focused tasks such as identifying and classifying SOA and other protocol information while providing greater control over how information is processed. Jarvis is actively exploring and developing SLM-based approaches for these focused clinical tasks.
LLMs provide valuable language understanding and reasoning capabilities and can be used selectively where those capabilities provide meaningful value. Jarvis does not depend on exposing an entire clinical protocol to a public LLM in order to perform its core workflow.
One important role for LLM technology within Jarvis is understanding human instructions. A Data Manager may request a field change, add a data collection item, modify a codelist, or provide broader CRF feedback using ordinary language — either directly in Jarvis or through an exported PDF casebook.
The language model’s role is to interpret the user’s intent. Jarvis then translates that intent into defined design actions that can be applied to the structured CRF model. This provides a natural interface while maintaining control over how the underlying design is changed.
As sponsors and CROs increasingly deploy approved private LLM environments, Jarvis is designed to take advantage of those capabilities where appropriate. This can enable broader use of AI on protocol information while maintaining organizational security and governance requirements.
Protocol Privacy
Control How Your Protocol Is Used
Clinical protocols contain valuable and often confidential intellectual property. Jarvis is designed to minimize exposure of protocol information to public LLMs.
Core protocol processing does not require sending the complete protocol to a public LLM. In selected workflows, limited protocol content — such as relevant SOA information — may be processed using LLM capabilities when enabled. Organizations can choose configurations that prevent protocol information from being exposed to public LLMs.
Where an organization provides an approved private LLM environment, Jarvis can be designed to make greater use of that environment according to the organization’s security and governance policies.
“Your AI strategy should adapt to your security requirements — not the other way around.”
Human-in-the-Loop
Automation Doesn’t Mean Autonomous Decision-Making
Jarvis helps identify, structure, generate, interpret, and update. Clinical professionals evaluate, decide, and approve.
The final authority remains with the clinical professional.
Organizational Intelligence
Turn Institutional Knowledge Into a Reusable Asset
Clinical organizations accumulate valuable knowledge over years of study development: standards, CRF libraries, historical designs, sponsor conventions, design patterns, and operational practices. Jarvis can incorporate this knowledge into the EDC design workflow so that organizational experience can be reused rather than rediscovered study by study.
Clinical Automation Built for the Real World
Clinical Automation Built for the Real World
See how Jarvis combines clinical knowledge, intelligent automation, AI, and human expertise.