Executive Summary
Inclusive+ aimed to elevate its culturally sensitive healthcare directory by developing an intelligent, AI-driven provider matching system. Leveraging Amazon Bedrock alongside a suite of AWS services, the platform delivers personalized, inclusive, and conversational search experiences tailored to diverse user needs. This end-to-end solution ensures low latency and high-accuracy recommendations while adhering rigorously to AWS security and compliance best practices.
Challenges
- Limited Personalization: The original directory search lacked tailored results, leading to generic recommendations.
- Missing Conversational Flow: Users couldn’t ask clarifying questions or refine their preferences interactively.
- High Sensitivity: Working with LGBTQIA+ health data required robust security and cultural sensitivity.
- Data Management: Integrating real-time user data with historical provider data required scalable, resilient data pipelines.
Scope of the Project
The project covered the end-to-end design, implementation, and operationalization of an AI-powered, culturally sensitive provider matching system for Inclusive+. It involved assessing platform limitations, creating a scalable, serverless AWS architecture (using Bedrock, Lambda, API Gateway, SQS, RDS, OpenSearch, Glue, and AWS security services), and integrating Retrieval-Augmented Generation (RAG) for real-time, inclusive recommendations. Key activities included developing conversational search flows with prompt engineering, building automated data pipelines for continuous data accuracy, and implementing real-time monitoring and security best practices such as KMS encryption, IAM least-privilege access, and Bedrock Guardrails. The project concluded with delivering a comprehensive runbook and knowledge transfer to enable the Inclusive+ team to confidently maintain and scale the solution, ensuring inclusivity and continuous improvement.
Partner Solution
We designed a fully serverless, secure AWS-native architecture leveraging Amazon Bedrock, Lambda, API Gateway, SQS, RDS, OpenSearch, and Glue. By combining Retrieval-Augmented Generation (RAG) with advanced prompt engineering and Bedrock’s generative AI, we created a conversational search experience that is culturally competent and inclusive. Data pipelines powered by Glue ensure fresh, accurate provider data, while monitoring and observability layers (CloudWatch, CloudTrail) ensure operational excellence.
Solution Architecture Diagram
Our Solution
- User Authentication & Access
- Amazon Cognito authenticates users securely, managing user identity and access to the personalized search experience.
- IAM roles enforce least-privilege access across all AWS services, maintaining strict security and governance.
- API Orchestration & Event-Driven Compute
- Amazon API Gateway acts as the secure entry point for user requests from the frontend, validating the identity token from Cognito and routing requests to AWS Lambda functions that orchestrate provider search and recommendation workflows.
- AWS Lambda(Initial Invocation) is responsible for initial request handling, where it validates the input, sends the query to Amazon SQS for async processing, and optionally calls Amazon Personalize to prefetch user recommendations.
- AWS Lambda (Bedrock Invocation) : This Lambda worker consumes messages from SQS, retrieves relevant provider data, and dynamically builds prompts that are sent to Amazon Bedrock Agents to generate a conversational and personalized provider response for the user.
- Amazon SQS decouples tasks, buffering search and enrichment workloads to ensure reliable scaling and fault-tolerant message processing, with downstream Lambda functions polling the queue to process provider matching jobs in parallel.
- Data Storage & Enrichment
- Amazon RDS (Provider Mapping ): Hosts structured provider profiles including identifiers, qualifications, and tags, serving as a source-of-truth that feeds into the Glue-based ETL pipeline and supports accurate retrieval and provider data linking.
- Amazon OpenSearch Service (Offers Index) : Indexes structured provider data such as specialties, inclusive practices, and locations, enabling real-time filtered search results that can be embedded into prompts or used directly for matching logic.
- Amazon S3 : Acts as a durable and scalable storage layer for unstructured provider data such as bios, documents, and metadata, which are accessed by AWS Glue for ETL processing and may be referenced in Bedrock responses or OpenSearch indexes.
- AWS Glue : Extracts and transforms provider data from S3 and RDS, formats and chunks it for embedding and indexing, and catalogs it for querying, with outputs feeding directly into Bedrock Knowledge Bases and OpenSearch indexes to ensure GenAI is operating on fresh, clean data.
- Generative AI & Personalization
- Amazon Bedrock : Stores indexed provider information in a Knowledge Base, processes natural language queries using Bedrock Agents, and enforces safety and inclusivity via Guardrails to ensure the AI responses are accurate, empathetic, and aligned with LGBTQIA+ values.
- Prompt engineering ensures that every response is contextually relevant and reflects the nuances of LGBTQIA+ healthcare needs.
- Retrieval-Augmented Generation (RAG) integrates real-time data from RDS and OpenSearch into Bedrock prompts, delivering dynamic, inclusive recommendations.
- Amazon Personalize (Provider Segmentation) : Analyzes user behavior and historical data to provide tailored provider recommendations, which are combined with GenAI results to further personalize suggestions based on click patterns, preferences, and feedback.
- Monitoring, Observability, and Security
- Amazon CloudWatch continuously tracks system performance, including Lambda errors, API latencies, and SQS queue depth, to ensure system health and reliability. It captures logs, metrics, and alarms across all services providing full observability, error tracking, and performance insights.
- AWS CloudTrail logs all API activity for auditing and compliance purposes, ensuring a full record of interactions and changes.
- KMS encryption secures data at rest across RDS, OpenSearch, and S3, while TLS encryption ensures data in transit remains protected.
- Bedrock Guardrails provide an additional layer of protection, filtering out unsafe, biased, or culturally insensitive outputs to maintain trust and inclusivity.
Benefits
- Personalized Matching: Tailored recommendations aligned with user preferences and cultural needs.
- Conversational Flow: Interactive, inclusive conversations that guide users to the right care.
- Operational Efficiency: Serverless components (Lambda, API Gateway) ensure low latency and cost efficiency.
- Resilient Data Flows: Automated Glue pipelines and robust monitoring prevent data gaps or mismatches.
- Security & Compliance: KMS encryption, IAM best practices, and Bedrock Guardrails ensure privacy and trust.
Metrics Used to Measure Success & Lessons Learned
To evaluate the effectiveness of the Inclusive+ Generative AI platform and ensure continuous improvement, we tracked several key performance indicators (KPIs):
- Relevance and Accuracy: Regularly measured the accuracy of AI-generated provider matches against verified RDS data and user expectations, ensuring personalized and culturally competent results.
- User Engagement: Monitored click-through rates, interaction depth, and return visits to gauge user satisfaction and trust in the platform.
- Response Latency: Targeted sub-second response times for Bedrock-generated answers, ensuring a seamless, real-time conversational experience.
- Error Rates: Tracked Lambda errors, API Gateway 4xx/5xx responses, and SQS queue backlogs through CloudWatch dashboards to maintain system stability.
- User Feedback: Captured explicit feedback through in-app ratings and qualitative surveys to guide prompt and response refinements.
- Cost Efficiency: Analyzed Bedrock token usage and Lambda/SQS metrics to maintain cost-effective scalability.
- Bias and Inclusivity: Conducted periodic reviews to ensure recommendations remain unbiased and culturally affirming for LGBTQIA+ users.
From these metrics, key lessons learned included the importance of iteratively refining prompt structures to align with evolving user needs and conversational nuances. Real-time feedback loops also highlighted the value of continuous prompt optimization and proactive monitoring of data integrity to sustain trust and engagement. These insights have driven ongoing enhancements, ensuring that Inclusive+ remains a leader in inclusive, culturally competent healthcare recommendations.
Outcome (Business Impact)
- Enhanced user satisfaction and trust through personalized, culturally sensitive provider recommendations tailored to LGBTQIA+ identities and healthcare needs.
- Improved match accuracy by 40% compared to the previous static directory search, driven by dynamic Bedrock-powered conversations and RAG-enhanced relevance.
- Reduced search-to-decision time by 35%, enabling users to find culturally competent providers more efficiently and confidently.
- Strengthened brand trust, with 92% of surveyed users indicating they felt more seen and supported by the culturally sensitive, inclusive platform.
- Eliminated manual intervention for prompt and data updates by leveraging fully automated data pipelines and real-time observability.
Executive Summary
BeNotable is a platform dedicated to connecting music students with colleges. To stay ahead in a competitive landscape, BeNotable aimed to leverage Generative AI to enrich students’ audition experience and differentiate their service. We assessed their existing AWS-based data infrastructure (Amazon S3 and DynamoDB), technical maturity, and business objectives. The assessment highlighted an opportunity to introduce the “Aria Audition Lab Coach”, giving students instant, AI-generated feedback on tone, rhythm, and expressive quality. This case study outlines how we implemented a secure, scalable, and cost effective Generative AI workflow on AWS.
Challenges
- Provide high quality AI feedback on large volumes of audio while maintaining low latency.
- Protect student data and intellectual property with robust security controls.
- Ensure end to end observability and graceful failure handling across asynchronous workloads.
- Integrate seamlessly with BeNotable’s existing AWS foundations without disrupting live users.
Scope of the project
- Discovery & Readiness : Assessed data quality, security posture, and AI objectives.
- Architecture & PoC :Designed an event driven, serverless architecture and validated model choice in Amazon Bedrock.
- Implementation : Built secure upload, processing pipeline, AI inference, and feedback delivery using API Gateway, Lambda, S3, DynamoDB, SQS/SNS, and EventBridge.
- UAT & Launch : Performance, security, and user acceptance testing with staged rollout.
- Enablement:Delivered IaC templates, runbooks, and a roadmap for multilingual expansion.
Partner Solution
- Cloud native platform - That matches music students with colleges via audition submissions.
- Web and chatbot interfaces - For students to upload recordings and receive feedback.
- Existing AWS foundations - Amazon S3 for raw audio and Amazon DynamoDB for metadata storage.
- Key business goals - Deepen student engagement, enrich learning experience, and stand out from competitor platforms.
- Secure Upload – Students authenticate with Amazon Cognito; requests are filtered through AWS WAF and served via Amazon API Gateway to a “PUT /upload audio” Lambda function.
- Storage Layer – Raw recordings land in an Amazon S3 bucket; Lambda captures metadata (student, instrument, timestamp) and writes to Amazon DynamoDB.
- Processing Pipeline – An SQS queue triggers a processor Lambda that transcribes audio and invokes Amazon Bedrock (Anthropic Claude or AI21) to generate feedback. Events are coordinated with Amazon EventBridge.
- Messaging Layer – Results are published through Amazon SNS. A Dead Letter Queue retains failed messages for replay and root cause analysis.
- Observability & Monitoring – Amazon CloudWatch Logs, metrics, and AWS X Ray traces provided full visibility, while AWS Config & IAM manage compliance and least privilege access.
- Scalability & Resilience – The design is serverless and fully managed, automatically scaling with usage and isolating faults through queue based decoupling.
Solution Architecture Diagram
Metrics Used to Measure Success & Lessons Learned
- Engagement: +30 % increase in average session duration; 2× rise in audition uploads.
- Latency: p95 feedback delivery < 4 s.
- Reliability: < 0.2 % message failure, all captured in DLQ
- Cost Efficiency: ~40 % reduction in operational overhead via serverless pay per use.
Lesson Learned
- Prompt engineering with few shot and chain of thought examples is key to nuanced music feedback.
- RAG with Titan Embeddings grounds generative output in music theory references for factual accuracy.
- Comprehensive observability accelerates latency tuning and error resolution.
- Early educator feedback loops refine model prompts and sustain content authenticity.
Outcome (Business Impact)
- Students receive immediate, high quality feedback, increasing practice frequency and quality.
- Colleges gain richer audition insights, improving talent fit decisions and placement rates.
- BeNotable differentiates as an AI driven innovator, attracting new users and institutional partners.
- Serverless architecture scales elastically with peak audition seasons while aligning costs to usage.
.png)
Customer Profile
Klamath Health Partnerships provides accessible, culturally sensitive, affordable, quality-driven, responsive, patient-centered health services to the community, with an emphasis on those who need it most. These underserved populations typically include individuals from low-income families, the elderly, disabled, children, mentally ill, developmentally disabled,immigrants, the working poor, and those unable to work. Klamath Health Partnership is the second largest medical provider in the region, offering residents access to culturally appropriate, high quality, and affordable primary and preventive health care from a network of local clinics.
Customer Pain Points
Klamath Health’s on-premise infrastructure and data center are on an active fault line, so secure backup and disaster recovery was a primary concern. They also wanted to migrate their EHR and other data to AWS, using Tableau Cloud for business intelligence and analytics. Their managed services provider, who had begun the planning process for a migration, was removed and they were left understaffed and questioning their plan. Additional factors included:
- Legacy software and infrastructure
- Data retention standards
- Structured file share permissions
- Antivirus, malware protection, anti-phishing
- Data auditing
- Security
- HIPAA compliance
Cloudtech Solution:
Cloudtech began the engagement with a one-day workshop to capture the desired technical and business outcomes.Importantly, Klamath Health leaders were present and active to provide a holistic view of what the organization needed and wanted from their architecture. Based on the identified goals, Cloudtech proposed technical solution options and, together with Klamath Health, prioritized the work required and created a detailed roadmap.
Following the roadmap, the Cloudtech team built an AWS presence, comprising Organizational, Security, Infrastructure, and Workload OUs managed through AWS Organizations.
Beyond account creation and AWS Control Tower implementation, the heavy lifting of the engagement was the data migration and hydration of the Klamath Health data lake housed on S3. To accomplish the requirements of a hybrid system with file share capability and multiple data sources of both structured and unstructured data, the team used Amazon S3 File Gateway, AWS Site to Site VPN, AWS Storage Gateway, and Amazon S3.
Throughout the engagement, Cloudtech provided continuous knowledge share and extensive hands-on training and advisory services to the Klamath Health team, instilling confidence to operate the solution going forward.
Customer Outcome
With this solution in place, Klamath Health has a HIPAA-compliant, resilient data lake to house their EHR data as well as otherbusiness-critical datasets. They also have a structured data lifecycle and a highly available data store to ensure the securityand integrity of their patient and physician data as well as a backup policy that meets their RPO and RTO requirements. Withthe balance of the cost savings of the removal of their managed services provider and the dialed in TCO of their new solution,Klamath Health is saving 77% on their infrastructure costsYoY. On top of that, Cloudtech has trained their team to be fullycapable of maintaining and growing their infrastructure as their patient base grows and their organization expands.
Solution Architecture Diagram
AWS Services Used:
- AWS Control Tower
- AWS IAM Identity Center
- AWS Cost Explorer
- Cost & Usage Report
- AWS Organizations
- AWS Directory Service
- Amazon CloudWatch
- AWS CloudTrail
- AWS Config
- AWS Key Management Service
- AWS Security Hub
- Amazon Macie
- Amazon S3 File Gateway
- AWS Site to Site VPN
- AWS Storage Gateway
- Amazon S3
Testimonial
.jpeg)
Project Summary
PXL is an open source social network platform for content creators. It enables users to create public or private spaces for any use such as any particular task base space or any other. Furthermore, users can take advantage of social features such as building connections, posting projects that pique the interest of other users, adding team members, notifications, project participation, and more. They can also manage their profiles and conduct a global search. This social network tool offers an online version where anyone can experience this free tool. PXL’s user interface is very logical, and users can easily navigate through various elements.
Problem Statement
The client’s requirement was to build a full-fledged backend application that can easily integrate with their prebuilt front-end application, and he later asked us to integrate the backend with the front-end.
We had to design and create a social platform where users can showcase their inventions and gain exposure. One can post any software project, categorize them, invite team members, and also participate in other projects.
Additionally, to meet the need for significant content uploads, a solution had to be developed that could easily handle the upload of media files while still being affordable and effective.
We also had to create a real-time notification system that monitors all network activity such as accepting requests, declining requests, and being removed from one’s network.
Our Solution
- With thorough testing, responsive design, and increased efficiency and performance, we concentrated on completing each task as effectively as we could.
- Based on the client’s requirements, we used S3 bucket, RDS, EC2, and flask microservice for media files and SES for emails.
- Amazon S3 was used for file hosting and data persistence.
– Amazon Relational Database Service (RDS) was used for database deployment as it simplifies the creation, operation, management, and scaling of relational databases.
– Amazon EC2 was used for code deployment because it offers a simple web service interface for scalable application deployment. - We sent emails using Amazon SES because it is a simple and cost-effective way to send and receive emails using your own email addresses and domains.
- Django-graphQL was used for the backend, and Next.js was used for the front end. Django includes a built-in object-relational mapping layer (ORM) for interacting with application data from various relational databases.
– GraphQL aims to automate backend APIs by providing type-strict query language and a single API Endpoint where you can query all information that you need and trigger mutations to send data to the backend.
– Next.js offers the best server-side rendering and static website development solutions. We utilized the flask microservice to help with high content uploads since flask upload files give your application the flexibility and efficiency to manage file uploading and serving. - Using Github’s automated CI/CD pipeline we have triggers for code lookup and deployment.
Technologies
Django-GraphQL, Next.js, PostgreSQL, AWS S3, EC2, SES and RDS
Success Metrics
- All deliverables were completed on time and exceeded expectations.
- Met all the expectations of the client and with positive feedback.
- The client was constantly updated on the status of the project.
Get started on your cloud modernization journey today!
Let Cloudtech build a modern AWS infrastructure that’s right for your business.