
Weekend Creative Challenge: Reimagining Kurdish Storytelling With Dengbej AI
Dengbej AI is a serverless storytelling platform inspired by the Kurdish dengbêj tradition. Originally introduced as an AIdeas project and selected as a semi-finalist, it evolved into a working prototype during the AWS Builder Weekend Creative Challenge. Built with AWS Lambda, Amazon Bedrock, DynamoDB, EventBridge, S3, and Amplify, the app curates news, generates Kurmanji narration, and presents stories through a radio-inspired interface.
Earlier this year, I published Dengbej AI: Kurdish Storytelling With Generative AI as part of the AWS 10,000 AI Ideas Challenge.
To my surprise, the idea was selected as a semi-finalist.
At that point, I thought I would immediately continue building it. But life — and other projects — had different plans.
I temporarily postponed Dengbej AI and shifted my focus to my small e-commerce business.
From Rugs to AI Commerce
Over the following months, the new AWS Builder weekly challenges inspired me to focus on a completely different problem.
Instead of building a storytelling platform, I started building AI tools for my small business.
Those challenges led me to create two new projects:
- RojKilim , my handmade rug and cultural heritage business
- RojAI , an AI-powered ecosystem that includes:
- An SEO-focused web application
- A Shopify AI agent designed for the Shopify Marketplace
Those projects became my primary focus.
I spent most of my time building tools that could help small sellers automate product creation, optimize listings, and improve their online stores.
The AWS Builder challenges became a regular source of inspiration, and they eventually led to something unexpected.
I earned two AWS Builder jackets while building those projects, but I still couldn't stop thinking about Dengbej AI.
Returning to an Unfinished Idea
This week's Weekend Creative Challenge gave me another opportunity to revisit the project that started it all.
The truth is that Dengbej AI never completely disappeared.
I had already built parts of the infrastructure.
Some backend services already existed.
The basic architecture was already there.
But the application was incomplete.
Instead of starting from scratch, I decided to take the work I had already done, finish the missing pieces, and finally publish a working prototype.
The Biggest Challenge Wasn't Technical
One of the biggest questions I faced wasn't about AWS.
It wasn't about AI.
And it wasn't about infrastructure.
It was a legal and ethical question.
How could I build an AI-powered news and storytelling application while respecting publishers?
How could I preserve attribution?
How could I avoid simply copying and republishing someone else's work?
Those questions forced me to rethink the entire project.
Instead of building another news aggregation website, I started building something different.
I started building a storytelling platform inspired by the Kurdish dengbêj tradition.
From News Aggregation to Storytelling
Traditionally, dengbêjs preserved stories, songs, history, and collective memory through oral storytelling.
They carried culture across generations.
That tradition became the foundation of this project.
The current version of Dengbej AI doesn't try to replace journalism.
Instead, external news sources are used for discovery and attribution.
The prototype:
- Ingests stories from RSS feeds
- Curates stories automatically
- Organizes stories into regional programs
- Generates Kurdish summaries
- Creates AI-generated Kurmanji narration
- Preserves source attribution
- Links readers directly to the original publishers
The goal isn't to reproduce articles.
The goal is to create a new storytelling experience inspired by the Kurdish dengbêj tradition.
Building a Serverless Storytelling Platform With AWS
One of my goals was to keep the architecture simple, scalable, and affordable.
The current prototype uses:
- AWS Lambda for ingestion, classification, processing, and APIs
- Amazon DynamoDB for articles, programs, events, and briefings
- Amazon EventBridge for scheduling and automation
- Amazon Bedrock for Kurdish summaries and narration generation
- Amazon S3 for media storage
- AWS Amplify for frontend hosting
The architecture looks like this:
RSS Sources ↓AWS Lambda ↓Amazon DynamoDB ↓Story Classification ↓Amazon Bedrock ↓Kurmanji Narration ↓News API ↓AWS Amplify ↓Dengbej AIOne of my favorite implementation details is the fingerprint-based pipeline.
If the underlying stories haven't changed, the system doesn't regenerate content.
That reduces unnecessary AI calls, improves efficiency, and helps keep operating costs low.
The First Working Prototype
Today, Dengbej AI can:
- Automatically ingest news from multiple sources
- Curate daily stories
- Group stories into regional programs
- Generate Kurmanji narration scripts
- Present stories through a radio-inspired interface
- Switch between English and Kurdish
- Preserve source attribution through direct links to publishers
The current programs include:
- Kurdistan
- Bakur
- Rojava
- Başûr
- Rojhilat
- Turkey
- Middle East
- World
One design decision I particularly enjoyed was allowing programs to adapt to the available content.
Some programs may contain five stories.
Others may contain only one.
Some may have no stories at all.
That flexibility makes the experience feel much closer to an editorial broadcast than a traditional news feed.
I'm Still Exploring the Future of Dengbej AI
The AI Ideas challenge gave me the original vision.
The AWS Builder weekly challenges gave me the opportunity to build RojAI.
And this Weekend Creative Challenge gave me the motivation to finally publish the work I had already completed on Dengbej AI.
I'm still exploring how this platform should evolve.
But one thing is clear.
I no longer see Dengbej AI as simply a news application.
I see it becoming a community-driven Kurdish storytelling platform.
I want to explore:
- AI-generated audio
- Complete Kurdish localization
- Community contributions
- Local journalism
- Oral history preservation
- Geographic storytelling
The prototype isn't finished.
But it's finally real.
And that's why I wanted to share it.
Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article