The problem Tagster solves

Long time ago, in the forests of ancient India, there lived a wise woman named Mata Shabri. Her dedication to serving Lord Rama was unmatched, and her efforts in gathering the best, most delectable fruits to present to Him became legendary. She would carefully choose the ripest, sweetest fruits, ensuring that only the finest offerings were given. This story of devotion and attention to detail has inspired Tagster – a platform that brings modern-day “Shabris” to the world of AI development, ensuring the highest quality datasets for machine learning models.

The Problem:

As AI and machine learning systems evolve, they need vast amounts of high-quality, reliable data to improve their models. However, obtaining clean and accurately labeled datasets has always been a critical challenge. AI companies struggle to gather vast amounts of data that are both consistent and high-quality. Traditional methods of gathering labeled data are slow, labor-intensive, and prone to errors. This is where Tapster steps in.

Tapster transforms dataset creation into a fun and gamified experience. Much like how Mata Shabri carefully selected the best fruits for Lord Rama, Tapster carefully curates data for AI models. Users engage with short-form content, such as images, short videos, and games. By selecting between pairs of AI-generated images, they generate valuable, labeled data that AI companies use to train better, more reliable models.

Challenges I ran into

I loved building the idea for this hackathon .. Major hurdle was that I am new to zk and going through documentation of zkverify was not easy.. Its not seamless. i faced many problems implementing it. It was facing lots of errors while compiling my circom, sometime r1cs was not generating, sometimes wasm was not generating etc.
I faced lots of issues while integrating these stuffs in js for proof generation.