The problem techTHELLA solves

When vendors go selling their products down the street, they are pretty ambiguous about where they will be able to make good sales. We solve this problem by predicting the chances of making the sale of their product in their proximity in the form of probabilities. For this, we have built an ML Model which can help us to predict these probabilities. A dataset of 1.25 Lakh entries is used to train this model. So the pain point of vendors is faced over here. So, a monthly subscription of Rs.150 is required to be bought by the vendor. Now, to predict we need certain data points and for that, we need data from the customers. To achieve this it is pretty clear that we need to govern the latest transaction that is made between the customer and the vendor but at the same time, it is not very customer friendly and apt to govern the payments. So , instead, we have used a tweaked approach in which on every sign-up by the vendor, he/she is assigned a QR code. Whenever a customer scans the QR code, the customer is enabled to provide a review to the vendor. This review helps us to update our algorithms to predict the probabilities as we know two things now 1> How many days veggies stay in customers' house 2> when they have made a purchase; which are enough for us to predict the next day of purchase helping us to predict for the vendor.
To gamify everything, whenever a customer provides a review we credit some points to the customer's account which can be redeemed in exchange for coupons; similarly, we use NLP protocols to check if the reviews are positive or not, accordingly we credit vending points to vendor accounts. Leaderboards are created for both separately.
To decrease the steps in the listing of items by the vendor we plan to use image processing by AI to do the same.

Challenges we ran into

Specifically, it was really tough because dealing with the newest languages was a bit tough.
NLP protocols had tough documentation.
ML model training needed to be conducted with a great data set
TensorFlow APIs couldn't be integrated because we ran it on colabs and professional use was forbidden in the free version.
Google Map Location APIs couldn't help us with exact mapping because of the free version.
Still, we have put together the most we could.