Inspiration

AI-powered stream finder.

What it does

Take a selfie with your family or group of friends and a stream will be matched to you matching the constraints of age range and gender. Take the decision of what you will see while taking into account the compromising factor in a novel mathematical constraint way.

How we built it

Infrastructure

  • We used pretrained aws rekognition models for face, age, and gender detection.
  • We also use our machine learning model in the back-end.
  • The whole solution is deployed with Streamlit.

Machine Learning

For the ML part mainly panda and sklearn: first merge the database you gave with the imdb database to get scores depending on user gender and age; then transform the text data either with one-hot encoding for classifying text input/ cosine similarity for nearest neighbors search. Then constraints on the search space via feature selection via xgboost, the standard techniques in recommender systems; more precisely: TF-IDF + SVD for text input and cosine similarity for nearest neighbors search

Business

we kept our decision along the way supervised by the core values we decided upon maintaining, which are building a product that maintains high-quality innovative tech solutions, and making it a profitable product and can impact the market of stream recommendations

Challenges we ran into

Taking into account the costumer and the business idea till the end, sure it would be cool to have more features; but that precisely goes against the decision fatigue we are addressing!

Accomplishments that we're proud of

  • Introducing dual novelty and staying true to our concept till the end.
  • Novelty of the idea of a selfie in itself.
  • Novelty for the recommender system via CompromAise: instead of finetuning a standard recommender model using NLP/ Nearest neighbors we went to constraint optimization keeping the user in mind.
  • maintaining high-quality standards for the technical solution, moreover we analyzed how and why streamifAI is a valuable product from business perspective.

What we learned

Even at our small scale, we have the means to create innovation!

Usage

no need for any code downloading for repos cloning, go use our live application yourself.

What's next for StreamifAI

To be adopted in production for leading the world in the new era of stream recommendation :)

Share this project:

Updates