What inspired you
Considering the current situation almost all essential interactions within an organization have turned virtual and employees are bombarded with meeting links across various platforms making it difficult to keep a track of all of them and we felt a dire necessity to have an end to end solution which enables effective collaboration amongst employees while also helping them to plan, organize and schedule all your meetings. Also, since not every employee gets to know about fellow employees and usually in huge organizations, every single meeting will at least consist of 10-200 participants but still a lot of their employees who join the meeting end up not knowing about 80% of other individuals joining them. So, it's essential for individuals to get their fellow worker's roles, skills, and interests to create integrity and better culture in an organization.
What it does
Vmeet is an ML-based web platform which not only enables users to organize all their virtual meetings within an organization but more importantly helps them to get to know their fellow participants before they virtually start a meeting with them. Based on their interests it also recommends fun and insightful virtual meetings happening across the world to relieve the user from work-related stress while making sure that they also constantly upskill themselves by virtual collaboration across diverse groups that are alike in this fast-paced digital world. This would also help prevent a strange culture within an organization where employees have to spend much time getting to know each other. This platform enables convenient virtual collaboration, where:
• After logging in: The user would be able to see their own profile and edit it. See their meetings dialog, their connections within the organization, tasks completed.
• They can see upcoming meetings on their portal which includes both mandatory meetings(ex: Daily Stand up meeting) + optional AI recommended virtual meetings which empower users at different locations to collaboratively share knowledge, have fun and feel relieved from the stress caused by the current outbreak (ex: abstract card meeting).
• Recommended Events: If the user clicks on the photo of the upcoming events, then it will take the user to a place where they can see more details about an upcoming event and that page will also have recommended events for them.
• Join the Meeting: A begin button empowers the user to directly join a meeting which is starting in 30 mins for you to join the call. If the meeting is in 2 days then the begin button is disabled until you approach within 30 min time frame.
• See who is Joining: This is crucial because if employees get to know fellow employees' who are joining the meeting their roles, skills, interests and also can connect with them or ask for them just to get to know them because of the upcoming meeting.
• Users can also see attendee demographics. For example, one employee could be joining from California, other from New York, someone from Singapore, India or Africa.
• Within the user portal, the user has an option to host new meetings in zoom or hangouts and automatically send invites as well. If the user chooses to host the meeting then it will take the user to a form where they fill out the data, time and purpose of the meeting and choose the people within the organization whom the user wants to join their meeting. The new meeting created will immediately be added into upcoming meetings and these attendees will be notified with the link in the mail and also they'll be reminded through a text message which is sent few minutes before the scheduled meeting time.
How we built it
Meetup recommendation system: To build the model we first webscraped April month’s information about virtual meetups using the Meetup.com API. For each group, We queried the organizer ID, category tags, location, the topics covered by it, and its member count then built a data analysis pipeline using Python on AWS EC2. Then chose Factorization Machines model which is a supervised machine learning technique and works great with large sparse datasets. It uses both the feature interactions and the group features. To build the system, we divided the data into cross-validation (CV) and holdout set, trained the model using the CV set. A content based RS was used since we had to use the categorical features of each group.
Visit our Github repo to learn more about how we built it.
1) Styling the web pages was tedious but we never compromised as our main goal was user convenience. So, we were highly prioritizing user UI/UX. We had to create 5 fully responsive polished web pages from scratch, which itself took around 30 hours of continuous effort to craft them and to create an end-to-end working model. At one point, we were scared if we could even get to our goal before the deadline.
2) Building an ML model to give meetup recommendations based on user profile data and meetup API keywords
3) Backend consisted of the integration of many API's( Twilio, Sendgrid, google-talks, zoom, calendly, esri, meetup) together with the MongoDB Atlas. We had to manually create a lot of user-profiles and fetch the data.
4) Integrating front end and Ml model with the back end
5) We initially were a team of five, two of them quit at an early stage pointing out this project is heavily overshot. So three of us took it as a challenge to prove we could do it.
What we learned
- Using Twilio API
- Dealing with MongoDB
- combining the recommendation model with the front-end.
- Product planning
- Time management and our peak capabilities
- UiPath for transferring Files
Try It out