TANDEM

Tandem Logo
Mermaid Map of most Technologies Used
Friends Plan a trip to New York City with LuisGPT!
Team's social graph in LA

Tandem

Tandem leverages the power of our two technologies - LuisGPT and Mangodb - in order to ease the process of travel and networking.

What TANDEM Does

TANDEM, or the Travel Automated Network Discovery and Experience Management tool, utilizes LuisGPT and Mangodb to provide a visual interface of connections and a comprehensive itinerary for the future, and is supported by numerous social media apps from which it draws its information from. By simply typing his/her username, a user can continue to chat with their friends in a groupchat while data is continually sent, processed, and updated in the background. Every 5 minutes, while the chat is active, an updated itinerary is provided. With this also available on the dashboard, users can see if they have primary, secondary, or even tertiary connections with someone who has either been to their destination before or currently lives there! This not only eases the process of planning a trip/vacation/work endeavor, but gives the user the ability to reach out to people that they might know through other people and/or facilitate conversation with people they know directly with the purpose of getting to know someone else!

LuisGPT

Screen Shot 2023-04-23 at 8 36 52 AM

This virtual assistant/chatbot reads and continuously updates a Firebase database with messages in a group chat that discusses future events and sends a GPT call with all relevant information about travel. Additionally, SerpAPI's Google Search Api gives GPT up-to-date information about future events, special occasions, and best sightseeing locations during the given timeframe. By combining all of this information, a comprehensive, detailed itinerary is returned.

Mangodb

Screen Shot 2023-04-23 at 8 39 28 AM

Mangodb, which stands for Massively Automated Navigation and Gathering of Online Databases, is our massively parallel web crawling engine that we used to extract information from numerous APIs and web scrape multiple sources in order to get personal information on our users (for demonstration purposes, we used ourselves). This way, we were able to identify primary, secondary, and even tertiary connections stemming from a single source.

What We Used

We used a plethora of technologies. These include:

Modal
Twilio
Firebase
Cohere
Tailwind
FastAPI
Next.js
React
Python
GPT4/3.5
SerpAPI
Tweepy
EasyCron
InstagramSweeper API
RapidAPI
Auth0
Midjourney
React Flow
MaterialUI

Challenges We Ran Into

API Implementation

We experimented with and utilized at least a dozen different API's for different services throughout the 36 hours. It was incredibly tough to find scraping API's without rate limits. We ended up paying 10 dollars for the InstagramSweeper API because the limit was 50 calls (that became 10,000, though).

Integrating Different API Services Together to Consolidate Information

Documentation for almost every API service is limited to simple functions and nothing more. Finding the quickest way to extract information from differently formatted json files and understanding the syntactical differences between different API's was quite difficult.