-
-
Architecture
Syntrix AI: Redefining Prior Authorization Through Intelligent Automation
Inspiration
Even in a world where hospitals run on advanced EHRs and predictive analytics, one process still lags behind: prior authorization. What should be a quick, data-driven approval often turns into weeks of waiting, delayed treatments, and administrative frustration. Prior authorization has become one of healthcare’s biggest pain points, costing the system over $31 billion each year and stealing precious time from patient care.
Dr. Raja Muppidi (Pharm.D, MS Health Informatics) has spent over six years blending healthcare expertise with strong technical and AI skills. With experience in data analytics, programming, and applied machine learning, he has built intelligent healthcare solutions that improve how information flows between providers and payers. Working at the intersection of clinical data and automation, he saw firsthand how AI could not only analyze medical information but also power real-time decision-making and streamline approvals.
Lahari Sandepudi (MS Computer Science) brings a complementary background in AI development and intelligent automation. Having worked on machine learning pipelines and cloud-native agent architectures, she saw the opportunity to design a system where multiple AI agents could collaborate to extract information, validate evidence, and execute approvals seamlessly.
When our paths crossed, we realized we were tackling the same problem from two sides: Raja from within healthcare’s AI and data ecosystem, and Lahari from the frontier of intelligent automation.
Together, we built Syntrix AI, an end-to-end AI-powered Prior Authorization system that automates clinical data extraction, document validation, and payer communication using AWS-native agents. What once took weeks now happens in minutes, enabling faster decisions, reducing denials, and ultimately helping patients get the care they need without delay.
What It Does
Syntrix AI is an AI-powered prior authorization system that automates the entire PA workflow using 6 specialized AWS Lambda agents. It's like having a team of expert medical coders, authorization specialists, and payer representatives who work 24/7, never make mistakes, and continuously improve.
The Complete Workflow:
- Upload clinical notes → Extract ICD-10/CPT codes using Amazon Bedrock Nova Pro
- Validate codes against official databases → Check medical necessity with AI reasoning
- Verify supporting documents → Simulate payer decision-making
- Generate authorization decisions → Provide transparent explanations
Key Features:
- Instant Processing: Upload → Authorization in minutes, not weeks
- AI-Powered Validation: Intelligent code validation and medical necessity review
- Evidence Verification: Automated document checking and compliance validation
- Transparent Decisions: Clear explanations for approvals and denials
- HIPAA Compliant: Full encryption and audit trails
How We Built It
Architecture: Modern serverless architecture on AWS with 6 specialized Lambda functions
Tech Stack:
- Frontend: Next.js 16 + React 19 with Tailwind CSS 4
- Backend: AWS Lambda (Python 3.11) with 6 AI agents
- AI Layer: Amazon Bedrock Nova Pro + Bedrock AgentCore
- Storage: S3 (HIPAA-compliant) for encrypted documents
- Database: DynamoDB with point-in-time recovery
- Security: IAM least privilege, full encryption, HIPAA compliance
The 6 AI Agents:
- Extraction Agent - Extracts codes from clinical notes using Bedrock Nova Pro
- Code Validator - Validates ICD-10/CPT codes against official databases
- Evidence Checker - Verifies supporting documents against requirements
- Payer Simulator - AI-powered payer decision-making using Bedrock Nova Pro
- Case Manager - Orchestrates workflow and manages case data
- Analytics Agent - Provides insights and reporting capabilities
Challenges We Ran Into
- Multi-Agent Coordination: Seamless handoff across 6 serverless Lambdas
- AI Latency: Optimizing Bedrock response times for real-time processing
- HIPAA Compliance: Designing secure, auditable workflows under tight deadlines
- Clinical Accuracy: Mapping AI outputs to real-world payer requirements
Accomplishments That We're Proud Of
- Reduced PA turnaround from weeks to minutes (95%+ time reduction)
- Achieved 100% HIPAA compliance with full audit trails
- Built 6 specialized AI agents that work together seamlessly
- Created production-ready system with real-world clinical validation
- Enabled transparent AI decisions with explainable reasoning
What We Learned
- AI Orchestration: Combining LLM reasoning with deterministic workflows
- Healthcare Compliance: Operationalizing HIPAA in multi-agent systems
- Clinical Validation: Mapping AI outputs to real-world medical requirements
- Scalable Architecture: Building cost-effective serverless solutions
What's Next
Immediate Goals:
- Integrate with major EHRs (Epic, Cerner, NextGen) via FHIR APIs
- Expand to payer-specific policy logic for multi-insurer support
- Build denial trend analytics dashboards
Long-Term Vision:
- Full-scale deployment across hospitals and payers as a SaaS platform
- Integration with Amazon HealthLake and Comprehend Medical
- Establish Syntrix AI as the industry standard for AI-driven healthcare administration
Our Commitment
At Syntrix AI, we're not just building another AI project—we're redefining how healthcare operates. We believe technology should amplify clinicians, not burden them, and every patient deserves faster, fairer access to care.
Built with compassion. Powered by AWS. Driven by data.
Built With
- aes256encryption
- amazon-web-services
- amazonbedrock
- amazoncloudwatch
- amazondynamodb
- amazonnovaactsdk
- amazons3
- amazonsagemaker
- apigateway
- awscloudformation
- awsiam
- awslambda
- bedrockagentcore
- bluebutton2.0sandbox
- cmscoverageapi
- framermotion
- hl7davincipas
- next.js
- nihclinicaltablesapi
- python
- react
- strandssdk
- tailwindcss
- tls1.2
- typescript
Log in or sign up for Devpost to join the conversation.