FUTURUM AI
- Financial Services
- Institutional Credit
Healthcare, Private AI
AI development (multi-agent Private AI system)
- Pre-assessment
- EMR charting
- Ambient listening
- AWS (encrypted)
- Amazon Bedrock
- Audit logging
- eClinicalWorks integration
- Clinician review workflow
Table of Contents
ABOUT PROJECT
Vidan is an AI-powered video intelligence platform revolutionizing security, access control, and operational efficiency. Beyond basic surveillance features, it offers advanced tools like dynamic data tagging for flexible analytics. Developed in partnership with Liquid Technologies, Vidan enhances security and insights across industries from manufacturing to retail.
About the Project
That work added 45 to 60 minutes to a clinician’s day.
Public AI tools were never on the table. Futurum’s obligations under the HIPAA Security Rule required every system touching Protected Health Information to process and store that data inside the organization’s own encrypted infrastructure.
Liquid Technologies designed Futurum AI from that constraint outward. Three agents, one encrypted AWS environment, and a clinician approval step in front of every entry that reaches a patient record.
Project Objectives
- Reduce post-visit documentation time without changing how clinicians work
- Keep all Protected Health Information inside Futurum's encrypted AWS environment
- Hold clinician authority over every entry that reaches the patient record
- Standardize note structure across clinicians and visit types
- Build a compliance foundation that extends to new workflows without a rebuild
Possible Solution
Liquid Technologies partnered with Vidan to build an AI-driven platform that boosts security, reduces downtime, and delivers actionable video insights. With features like gun detection, fire alerts, and real-time data tagging, the system meets industry-specific needs. Its user-friendly design ensures effective use by both security and operations teams across environments.
The Problem: Documentation Overhead Under HIPAA Constraints
Three constraints shaped the build.
Documentation overhead.
Documentation overhead. Clinicians wrote notes from memory after visits. Record quality varied across clinicians and visit types, and entry work consumed 45 to 60 minutes per clinician per day.
EHR governance.
eClinicalWorks FHIR APIs do not populate chart fields from external inputs. Conditions, medications, and allergies need a structured, validated update path. We treated that as a design requirement rather than an obstacle. No system should write to a patient chart without clinician sign-off.
Data residency.
HIPAA governs every system that processes or stores PHI. All inference, handling, and storage had to happen inside Futurum's own encrypted infrastructure, which ruled out shared model APIs before design started. This is the condition that makes a private AI deployment model the only workable option.
The Solution: A Three-Agent Private AI System
Each agent owns one stage of the documentation workflow. All three run inside Futurum’s private AWS tenancy.
Ambient Listening Agent
Captures and transcribes the consultation in real time inside the private environment, extracting clinically relevant detail from the interaction. Output: a structured transcription.
Pre-assessment Agent
Reads the transcription against the patient’s existing record and drafts a structured pre-assessment summary with supporting notes. Output: a draft summary for clinician review.
EMR Charting Agent
Formats the approved summary into a chart update package for eClinicalWorks covering conditions, medications, allergies, and care decisions. Output: a validated package ready for clinician-authorized entry
How the Workflow Runs
No PHI exits Futurum’s private AWS environment at any point in that sequence.
Private AI Architecture on AWS
| Layer | How It Works |
|---|---|
Ambient |
Real-Time Audio Processing Inside The Private AWS |
AI Inference |
Amazon Bedrock In Stateless Configuration. No |
Foundation |
Anthropic Claude Via Bedrock, Deployed Inside |
Agent |
Each Agent Runs As A Versioned Role With Constrained |
Backend Control |
Every Data Interaction Routes Through Backend-Controlled |
Storage |
All Outputs And Session Data Held In Futurum's Encrypted |
Connecting AI to systems that were never built for it is where most healthcare deployments stall. The eClinicalWorks integration was designed around the API’s actual behavior rather than assuming it away.
Security, Governance, & HIPAA Alignment
PHI residency.
Patient data is processed and stored only inside Futurum's encrypted AWS environment. Nothing transmits to an external AI service
Permission scoping.
Agents hold constrained, versioned roles with no direct database access and no autonomous calls to eClinicalWorks
Two-stage clinician control.
Every agent output requires clinician review. Chart updates require a second validation before EHR entry
Manual reconciliation.
Clinicians reconcile conditions, medications, and allergies before submission
Full audit logging.
Every agent interaction, approval, edit, and rejection is logged with user ID, timestamp, and output version
Business Impact
| Before | After Futurum AI |
|---|---|
|
|
|
|
|
|
|
|
|
|
Private AI questions from financial services buyers
A private AI deployment runs inference inside the organization’s own cloud tenancy rather than calling an external API. In this build, Amazon Bedrock ran in stateless configuration inside Futurum’s encrypted AWS environment, so no Protected Health Information reached a third-party service at any stage.
It means no AI output reaches the patient record without human approval. Here, every draft summary is reviewed by a clinician, and every chart update package clears a second clinician validation before entry into eClinicalWorks.
A signed agreement covers the legal relationship. It does not change where data travels. Organizations with strict residency requirements need the processing itself to stay inside infrastructure they control.
Why Private AI Was the Only Viable Model
Why Private AI Was the Only Viable Model
Healthcare AI deployments fail in two predictable ways. They miss compliance requirements, or they bypass clinical governance and create accuracy and liability exposure.
Futurum AI started from both constraints instead of retrofitting controls onto a public tool. The three-agent structure maps to how documentation already works: listen, summarize, update the record. Each step is governed, each output is reviewed, and the clinician holds authority over everything that touches the patient record.
The compliance foundation now supports additional workflows without a rebuild.
See what a private AI deployment looks like in your environment
Tell us about your project and what you’re looking to build. A Liquid Technologies specialist will help you identify the right AI, data management, mobile app, or website development solution based on your goals, budget, and timeline.
- AI-powered solutions — Build intelligent AI models and solutions tailored to your business needs.
- Data-driven solutions — Manage, organize, and leverage your data to support smarter business decisions.
- End-to-end development — From mobile apps to websites, we build scalable digital experiences designed around your goals.
