We are building a better matching algorithm that would allow people to discover and foster meaningful connections in the communities, based on shared interests and values.
Today the internet has grown to be this vast community of resources, entertainment, economies (games), and bubbles. Your internet could be vastly different from my internet and it is largely categorized by how we consume the internet. This bubble creation has led to massive impacts on economies and macropolitical and economic trends as well. Unfortunately for the poor and underprivileged, the internet is very different as compared to the rich and well-off. This social divide brings up very nuanced differences in culture and is expected to accelerate in the next 20 to 25 years as data becomes cheaper and internet speeds become faster.
Another major problem that we have is that this shared data is in complete control of the web2-based organizations which often partner with entities that offer them the most money for our attention directly (ads shown on these platforms) or indirectly (recommendations based on algorithms) methods. This data ownership gives these companies massive valuations and revenue streams. Although the engineering required to build these platforms would be redundant without the user input, the users are not getting anything out of it and are dependent on the organization for everything in an unfair manner. There is very little transparency as to what these organizations do with our data.
The internet as a whole allows one to be anonymous and still consume, but establishing trust by being pseudo-anonymous is an extremely difficult task for economies. Especially for growing permissionless economies that have taken shape with play-to-earn/decentralized finance/decentralized autonomous organizations. These are expected to grow and would require an identity badge in order to evaluate the potential members and evolve the ecosystem.
We were faced by multiple technical challenges the most notable was that circom and mina contract had limitations as to how much data can be passed for processing, so sending full natural langague processed word embeddings (each vector being 50 N-Dimensional array was an a challenge in itself, we ended up resolving it by taking some parts of it on frontend so that we still are able to preserve the user privacy and not have any backend systems.