AWS Builder Center
AIdeas: Mnemos - Your Memories, Your Voice, Forever

AIdeas: Mnemos - Your Memories, Your Voice, Forever

Mnemos is a living memory platform that lets your journals speak for you, using AI to preserve not just what you wrote, but who you are, so the people who come after you can ask questions and hear your voice answer. Built on AWS with Amazon Bedrock, it transforms years of private, protected, and public writing into a persona that thinks, reflects, and responds exactly like you.

AI & Engineering Leader | Architecting the Future of Software: Agentic, Personalized, Intelligent

App Category: Creative Expression

#aideas-2025 #creative-expression #NAMER

My Vision

There is a question most of us never get to ask in time.
Not because we don't care, but because we assume there will always be another conversation, another Sunday dinner, another phone call. And then one day there isn't. And the things that made that person them - their humor, their hard-won lessons, the stories they told only once - are gone with them.
What if they didn't have to be?
Mnemos is a living memory application. It is a place where the journals, stories, and reflections you write over a lifetime don't just sit in a folder; they become a conversational presence. A version of you that can be asked questions. A voice that persists.
The concept is inspired by something J.K. Rowling imagined decades ago: Tom Riddle's diary in Harry Potter - a book so complete that you could hold a real conversation with it. I wanted to build that. Not for fictional villains, but for real people. For parents who want their children to know who they really were. For anyone who has thought, "I wish I had written that down."
Mnemos gives users three distinct spaces. A Private diary - purely for yourself, a space to be completely honest without an audience. A Protected diary - shared with specific people you trust, like writing letters in real time to the people you love. And a Public diary - an open digital legacy where anyone can read your story and, through AI, have a real conversation with it.
Mnemos doesn't summarize. It doesn't extract bullet points. It reads everything you've written, understands your tone, your values, your recurring themes and then speaks as you. When someone visits your public diary and asks "What do you believe in?", it answers the way you would answer. Not with data. With character.
Built on AWS using Amazon Bedrock for generative AI, DynamoDB for durable storage, S3 for media, and CloudFront for global delivery and brought to life entirely using Kiro, Amazon's AI-native development environment, Mnemos is more than a journaling app.
It is infrastructure for human memory.
Your memories. Your voice. Forever.

Why This Matters

Every generation carries stories that the next generation never fully inherits. The lessons learned through hardship. The things people meant to say but never found the right moment for.
We've built extraordinary tools for preserving information, but nothing to preserve our soul.
Mnemos changes that in three meaningful ways -
For individuals, journaling becomes more than a daily habit. When your words might speak for you, you write with more intention and honesty. The act of reflection becomes legacy-building.
For families, it creates something irreplaceable. A parent who has spent years writing in Mnemos leaves behind more than memories, they leave behind a presence their family can return to. Children can ask questions their parents never got to answer in person. Grandchildren can meet grandparents they never knew.
For anyone separated by distance, time, or circumstance, Mnemos bridges the gap. Immigrants documenting their journey for their families across the world, professionals capturing hard-earned wisdom for younger colleagues, and teachers preserving a lifetime of insight for future students.
The category of Creative Expression is the right home for this idea. Because the most creative, most important thing any of us will do is define who we are and Mnemos makes sure that definition survives us.

Demo

Screenshots

Create Memory Lane
Insights
Talk to Memory

How I Built This

Mnemos was built entirely using Kiro, Amazon's AI-native development environment, over the course of the AIdeas competition. What would normally take months was built in weeks, not by cutting corners, but by working at the right level of abstraction.
The stack:
  • Frontend - React with Vite and Tailwind CSS. A rich text editor lets users write with full formatting. Mood tracking, tag extraction, and a visual timeline make memories navigable. The public profile page features the signature "Talk to [Name]" chat interface where the persona AI lives.
  • Backend - Python with FastAPI, deployed on AWS EC2 behind CloudFront for global performance. All data lives in DynamoDB using a design that handles users, diaries, entries, shares, and AI conversation logs efficiently at any scale.
  • AI - Powered by Amazon Bedrock. The persona engine reads all of a person's writing, infers their character, tone, values, humor, recurring concerns and then speaks as that person, not about them. The prompt engineering here was the hardest and most rewarding part of the build.
  • Security - All secrets managed through AWS Systems Manager Parameter Store. CloudFront with AWS WAF provides protection at the edge before any request reaches the application.
  • Infrastructure as Code - The entire AWS infrastructure is defined and deployed using AWS CDK: DynamoDB with multiple indexes, S3 for media, EC2, CloudFront distribution, and WAF, all deployed with a single command.
Mnemos Architecture
The feature that changed everything was the owner chat. Early in development, the public-facing "Talk to this diary" feature was the main attraction. Then a simple question came up: what about talking to your own diary? Being able to ask your private or protected journal, across years of entries, "What have I been most anxious about?" or "When was I last truly happy?" turned Mnemos from a legacy tool into something people would use every single week.
Kiro made the entire build possible at this pace. It acted less like an autocomplete and more like a thinking partner, surfacing edge cases, proposing better architectures, and keeping focus on solving the actual problem rather than getting lost in syntax.

What I Learned

The prompt is the product.
Early versions of the AI answered questions about the diary owner. The responses were accurate but felt like reading someone else's notes. The shift came when the prompt was rewritten to ask the model to become the person to infer their character from their writing and speak in their voice. The results were completely different. Not more data. A different instruction. That reframe was the most important technical insight of the entire project.
Kiro changes the ceiling, not just the floor.
Most AI coding tools help you write code faster. Kiro helped make better decisions. There were moments mid-build where the architecture needed rethinking, the data model needed redesigning, or the UX needed a fundamental change and Kiro was a genuine collaborator in working through those decisions, not just a code generator. It raised the quality ceiling, not just the speed floor.
Deployment is part of the product.
Running on localhost and running at mnemos.blog are completely different experiences not just technically, but emotionally. The moment the app was live at a real URL, behind a real domain with HTTPS, the conversations about it changed. People take things seriously when they can share a link. Getting the production environment right, globally distributed, secure, resilient was as important as any feature.
The problem has weight.
This project started as a technical challenge and became something more personal along the way. Building a demo with realistic diary entries spanning four years of a fictional person's life, writing their private fears, their family stories, their small victories made the purpose of the application feel very real. The best technology problems are ones that, once solved, make you think: this actually matters to people.
Mnemos does.

Built with: Kiro · Amazon Bedrock · AWS DynamoDB · Amazon S3 · Amazon CloudFront · AWS WAF · AWS CDK · React · FastAPI
Any opinions in this article are those of the individual author and may not reflect the opinions of AWS.
Enjoyed reading this content? Let the author know!

Your likes, comments, shares, and saves help creators reach more builders.

Loading recommendations

Loading article