AIdeas: Transforming Healthcare into AI powered Wellness Companion

Managing health is tough. Symptoms shift daily, reports pile up, and vital details vanish between brief doctor visits. Patients struggle to remember, caregivers feel uncertain, and doctors decide with incomplete information. Mimamori unifies fragmented health moments into a complete story capturing voice logs, organizing medical documents, and delivering AI-powered alerts. Care becomes simple, continuous, and shared so nothing is overlooked.
Shafayat Saad

Published Feb 25, 2026
Last modified Mar 17, 2026
Every health moment matters, yet data is often fragmented and lost.
Mimamori captures your entire health story in minutes, connecting you with your loved ones. Care becomes simple, complete, and shared. So, nothing is ever overlooked.
﻿
App categorySocial Impact · Daily Life Enhancement
﻿
My visionMimamori (見守り — Japanese for "to watch over") is a full-stack, AI-powered health companion that transforms the messy reality of chronic illness into a clinical-grade data pipeline ,one that patients control and doctors trust.
It is not just a passive tracker; it is an active, reasoning agent that synthesizes daily patient logs and historic medical documents to generate actionable awareness.
We built Mimamori to bridge the gap between a patient's lived experience and clinical intervention. The platform features:
Intelligent Voice Logging: Patients can speak naturally about their symptoms ("I've felt unusually tired for three days and my legs are slightly swollen"). The AI parses this into structured medical metadata, no complex forms required.
Document Vault & Intelligence: Users can upload lab reports, prescriptions, and discharge summaries. Mimamori automatically extracts key entities to build a longitudinal health record.
Smart Alerts & Predictive Precaution Engine: Our core orchestration agent correlates daily logs with historical document data. It identifies subtle deviations (like a worsening trend indicative of a medication side effect) and alerts the patient and their Care Circle before the situation escalates.
Doctor-Ready Visit Preparation: With a single click, the platform generates a concise, one-page PDF summary detailing top recurring symptoms, adherence, and abnormal trends to make the 15-minute doctor consultation highly effective.
The Care Circle: A non-intrusive, view-only dashboard for family members to monitor adherence and receive critical alerts, providing support without surveillance.
﻿
Why this mattersIn the United States, six in ten adults live with at least one chronic condition (CDC, 2022). Despite this high prevalence, healthcare delivery remains largely episodic, relying on brief clinical consultations often limited to approximately 15 minutes to evaluate weeks or months of patient experience. This structural limitation creates what may be termed the “Clinical Data Gap”. The substantial volume of undocumented health information that accumulates between medical appointments.
During these intervals, symptoms fluctuate, lifestyle factors change, medications may be adjusted informally, and clinically significant patterns may emerge unnoticed. As a result, patients frequently rely on subjective memory when reporting their condition, caregivers experience uncertainty regarding day-to-day health status, and clinicians must make decisions based on incomplete data.
Mimamori addresses this gap by enabling continuous, structured documentation of daily health metrics, emotional states, prescriptions, and laboratory reports. Through AI-assisted analysis, the system transforms fragmented inputs into clinically meaningful insights, identifying patterns and correlations that support more informed decision-making.
In addition, the platform facilitates secure data sharing with designated caregivers, providing timely alerts when potentially serious changes are detected. By integrating longitudinal health tracking with intelligent analysis and controlled information sharing, Mimamori promotes a shift from reactive, episodic care toward a more continuous and proactive healthcare model.
Ultimately, it recognizes that health is shaped daily not solely within clinical settings and therefore requires continuous observation and structured interpretation.
﻿
How we built thisMimamori was built with an uncompromising focus on accessibility, security, low-latency intelligence.
The High-Level Architecture:Frontend / Orchestration: We deployed a responsive Next.js 14 application utilizing React, TailwindCSS, and Framer Motion for a fluid, accessible User Experience.
Database & Persistence: We implemented a hybrid approach. We use Amazon DynamoDB via aws-sdk for high-throughput, highly available storage of symptom logs and user authentication (MimamoriUsers, MimamoriData). We also utilize Prisma for structured relational management where necessary.
The Intelligence Layer:
Amazon Bedrock (Nova Micro & Pro): Serves as our "Reasoning Engine." In our 
medical-ai.js module, we use the lightweight Nova Micro for lightning-fast document classification and routing, while relying on heavier models for complex symptom correlation and generating the Doctor-Ready summary.
Amazon Comprehend Medical: Used via DetectEntitiesV2Command to extract precise medical ontologies (medications, conditions, vitals) from the colloquial text logs.
Amazon Textract (Integrated): For OCR and structured data extraction from uploaded lab reports and PDFs.
Storage & Notifications: Amazon S3 handles secure, encrypted document storage, while Amazon SES and Amazon SNS power our Smart Alerts and Care Circle notification workflows.
The "Predictive Insight" Loop: The technical breakthrough of the application was our serverless ingestion pipeline. When a patient submits a voice log, the audio is transcribed, the unstructured text is fired through Comprehend Medical into structured JSON, and Amazon Bedrock instantly cross-references the new entities against the historical dataset in DynamoDB. If a defined threshold is breached, the circuit breaker trips, instantly dispatching an SES/SNS alert to the Care Circle.
Demo ﻿
﻿
Here we can create an account as a patient or a caregiver. Caregivers require a 6-digit code from the patient to access only the details that the patient chooses to share.
﻿
This dashboard displays a summary of the health and medical information shared by the user, including metrics such as blood pressure, hydration levels,next dose details, and sleep patterns.
﻿
Users can record daily health data through text or voice on this page, including blood pressure, sleep patterns, and medication details, which are later transformed into insights on the dashboard.
﻿
The Health Trends page provides an analytical overview of the user’s recorded health data, showcasing key indicators such as blood pressure, sleep patterns, medication logs, energy levels, and pain tracking. It includes an overall wellness score, AI-based correlations, and insights derived from user inputs and uploaded medical documents.
﻿
Here,the user can view an upcoming follow-up consultation with a doctor and has the option to add, search, or prepare for appointments.
﻿
This page allows users to prepare for an upcoming doctor visit by reviewing recent symptoms and diary logs. Users can generate an AI draft summary, add notes or questions, and export the summary to discuss during their consultation.
﻿
This page allows users to securely store, organize, and manage their medical documents in a digital Document Vault. Users can upload new records, categorize files (such as lab results, prescriptions, or doctor notes), and view recent activity. It also shows storage usage and a complete list of uploaded files for easy access and management.
﻿
Users can click on a prescription, lab report, or clinical notes from the Document Vault to open and view detailed information. After selecting a file, they are taken to a page where the uploaded report is displayed and analyzed. Here, we can see that the clinical notes are clearly described, helping users better understand the medical details.
﻿
After uploading all information, here we can see a comprehensive summary on the dashboard along with all previously added details. After uploading all information, here we can see a comprehensive summary on the dashboard along with all previously added details.
﻿
The Add Caregiver page enables users to invite a caregiver, such as a family member or nurse, to assist in monitoring health data. It allows configuration of contact details, alert preferences, and access permissions, including daily logs, notifications, and document access.
﻿
This page allows users to send email alerts to caregivers based on recent health activities. Users can select relevant log entries, generate a draft message, add a custom note if needed, and send the alert instantly.
﻿
This is how Mimamori flags and sends automatic email flags as well as manual flags. 
﻿
How Kiro IDE Accelerated DevelopmentBuilding a complex, context-aware AI application requires rigorous design and bulletproof architecture. AWS Kiro IDE was the force multiplier that enabled me to build Mimamori from concept to production-ready MVP in record time.
"Architecture Before Code" Mastery: We utilized Kiro IDE's 
.kiro/masterplan.md to define our stringent AI interaction models, security boundaries, and data schemas before writing a single line of logic. Kiro scaffolded our Next.js frontend and Prisma/Better-SQLite3 backend, ensuring our early foundations were rock solid.
Premium UI Execution: Healthcare apps often look sterile and intimidating. I used Kiro to systematically apply modern, engaging aesthetics specifically implementing advanced Glassmorphism, fluid Framer Motion micro-animations, and a highly responsive Bento Grid layout. Kiro helped ensure our components felt premium, responsive, and deeply reassuring to the user.
Agentic Refactoring & Iteration: When integrating complex AWS architectures, Kiro served as an invaluable pair programmer. Whether debugging Next.js server actions, implementing the Hyperspeed hero animation, or structuring the DynamoDB authentication schemas, Kiro’s context-aware suggestions kept development moving at an aggressive pace without compromising code quality.
Cost break down ﻿
﻿
This expense summary highlights a total cost of $33.23, primarily driven by domain registration via Amazon Registrar ($15.00) and AI processing using Claude Opus 4.6 on Amazon Bedrock, where we first used 2,000 Kiro free credits and then $9.72 worth of Kiro credits inside the Kiro app to build the full application. The remaining balance includes AWS Amplify hosting ($4.72), taxes ($2.10), Claude 3.5 usage ($0.67) for in-app reasoning, and other small fees.
Future features:Social Alerts Integration: Extend notifications to WhatsApp, Messenger, and other platforms, ensuring caregivers and loved ones never miss critical updates.
Caregiver Full Access: A paid service that grants caregivers complete visibility into patient health stories, empowering them to provide proactive support. 
Pro Subscriber AI Caregiver: Subscribers gain the ability to chat directly with Mimamori’s AI asking questions, delegating tasks, and receiving personalized insights. It can do everything a human caregiver can, and more, making health management interactive, intelligent, and always available.
Doctor Collaboration Portal Secure: streamlined access for physicians to review patient logs, lab reports, and AI‑generated summaries before consultations.
On Platform Online Doctor Consultation: Built‑in telehealth functionality that allows patients to connect directly with medical professionals, share their unified health story, and receive timely guidance without leaving the app. 
In‑App Messaging Secure messaging between patients and caregivers, enabling real‑time communication, reassurance, and coordination without relying on external apps. 
What I learned"Small and Fast" beats "Big and Slow." Orchestrating Nova Micro for classification and Claude Haiku for reasoning created a snappier, more reliable experience than routing everything through a single large model. Each model does what it's best at.
Context transforms AI from generic to profound. By grounding Bedrock with the patient's Medical Baseline (their unique vitals targets), their historical logs, and real-time environmental data (weather), the AI produces correlations that feel genuinely insightful, not boilerplate.
Trust is earned through control. The "Human-in-the-Loop" pattern , letting patients edit AI-generated summaries and check/uncheck suggested doctor questions  is not just a UX feature. It's the single most important trust-building mechanism in the entire application. Patients who feel in control of their data use it more.
The invisible features matter most. Users don't notice that their hydration goal is quietly adjusted to 2400 mL because it's 34°C outside. They don't notice that the AI asked about dizziness today because they logged a headache yesterday. But these invisible, context-aware moments are what separate a "health tracker" from a "health companion."
kiro/masterplan.md first caught dozens of architectural edge-cases regarding data silos before they were ever coded. When building social impact apps, writing the requirements first is the most cost-effective engineering decision you can make.