KardsAI - Instant Flashcards Mobile App

KardsAI generates flashcards from any PDF, Text, or Prompt in seconds. Skip the tedious hours of preparing to learn and get right to studying - anything, anywhere, anytime.

Comment

Inspiration

The inspiration for KardsAI came from my personal experiences while traveling regularly through Spain, where I frequently faced the challenge of needing to know topic-specific vocabulary. An incident last year, where my car broke down with a flat tire, highlighted the need for a quick way to refresh my knowledge on specific topics to effectively communicate with locals. As a developer looking for a meaningful project in the AI space, this challenge sparked the idea for KardsAI.

What it does

KardsAI is a mobile app available on iOS and Android that allows users to generate flashcards on any topic using prompts, uploaded PDFs, or pasted text. It's designed to solve problems not only for lifelong learners and travellers but also for students or anyone looking to start studying quickly. By instantly transforming text or documents into flashcards, KardsAI eliminates the hours of study preparation, enabling users to begin learning any topic immediately.

It implements a spaced repetition algorithm with smart card notifications to decrease the time to learn while increasing the memory retention time.

How I built it

I started with a rough design outline from my prior skydiving logbook app Squifly and used the combination of flutter and firebase to start quickly. It took around 240 hours for the first version - which I am quite happy with speed wise. Its clearly noticeable that the new AI code helper accelerate the process quite a bit.

Challenges I ran into

The most difficult part of the project to nail down was the way to generate the flashcards consistently. After the initial setup was done, the API worked in general more and more issues popped up during testing. The API responses were inconsistent, it had problems with repeating content and duplication of content. Depending on the text it sometimes went into loops and end of last year the model laziness popped up as issue. Its quite clear that the model reliability of current systems is not quite there were it can be relied upon from developer perspective without taking extensive measures for quality control and latency handling. I assume this will be an ongoing process since AI models change and user input varies greatly both from input quality as well as expected outcome. Finally language consistency for a global app is also not easy to navigate. But in the end I am quite happy with the current level of quality and will improve upon that moving forward.

Accomplishments that I'm proud of

I am quite happy with the current level of flashcard quality that is being generated. It is genuinely an improvement for my every life to have at hand while travelling and I am surprised how well the timed notifications work for increasing my learning speed while keeping me from slacking off.

What I learned

I was surprised and partly shocked how difficult it is to handle the AI API backend calls. I had to jump to a lot of whoops to circumvent bad quality and I assume this will be an ongoing process. It will be very interesting to see how 2024 develops as I assume a lot of companies and products will have the same issues.

What's next for KardsAI - Instant Flashcards Mobile App

I am adding features constantly with the goal to make it the best AI first learning app on mobile - competing with the big legacy existing apps. User feedback is good and I am happy to have gotten quite a few ideas for enhancements already - from adding friends in app to share existing decks to bundling decks in collections. I hope to add more gamified content such as quizzes and motivational streaks to blend the level between learning and fun games more.

Built With

Share this project:

Updates