This is the demo of Marax's Intelligent Matching Platform focussed on the online dating use case. Its shows how the users are intelligently recommended matches based on their interaction with previous matches and other relevant features.

The app shows features which determine whether a user likes the profile or not. The features shown are Age, Friends of friends, Matching likes, height and interests and location. There is also an info button to show more information on the profile

Data Collection

Gaurav responds to user matches based on features displayed such as age, height, location and Interests, his likes and dislikes, which are retrieved from a facebook Login, and also his response to the profile is recorded. Thus, we now have a list of User Action data of Gaurav.

User ID Age Height Location Interests Matched ... Viewed Profile User Response Response back Messages Exchanged Sentiment Score
2112 23 189 Bangalore 1 ... 1 1 1 12 8
2312 24 154 Pune 0 ... 0 0 0 0
2783 20 180 Chennai 3 ... 1 1 0 0
1242 21 169 Delhi 1 ... 0 1 1 12 2
3466 22 167 Bangalore 0 ... 1 1 1 3 4

The User Action data consists of features which are visible on the app, like interests, Location etc., and it also contains data of Gaurav’s response to the profiles and whether he viewed the profile or not. The Response back is a positive value if Gaurav receives a response back from the profile he responded to. Messages shared feature is used to determine whether there is a genuine interaction between Gaurav and the users. Further more, the conversation can be judged as positive or negative in nature by using Natural Language Processing and performing a sentiment analysis on the conversation.

The response rate helps in determining the attractiveness level of Gaurav and can be used to recommend similar users who would find him attractive. Based on the responses, a popularity score can also be calculated for users so that the popular users are not shown to everyone. There are many other features which are used to increase the accuracy of the matches but cannot be shown in the demo.

Model Training

Marax’s AI Engine then feeds the data to multiple proprietary algorithms and aggregates the results to give out a list of best matches for Gaurav. Marax’s AI engine works very well in cases like these because there is a continuous influx of new user action data. The data is continuously fed into Marax’s AI engine to give out the best recommendations for the current state of the user. Thus, Gaurav is matched to people, who are attracted to him and also have features which Gaurav finds attractive. Similarly, the matches are recommended for other users, Jeff, Maria, Joseph, Jane, based on their interaction in the app. You can interact with a working model of our recommender system below.

Click on Jeff to get started.


  • Jeff Henry


  • Maria Lin


  • Joseph Tribbiani


  • Jane Green


  • Gaurav Sharma




  • Kelly Decosta


  • Emma Geller


  • Susan


  • Lisa


  • Michelle