CryptoWhisperer

demo 1
demo 2

Inspiration

For being a smart and up-to-date crypto trader, one must need to catch up the trending crypto market news so that (s)he can always make a wise action to in their trading strategies for better profits. CoinDesk is one of the famous crypto sites where traders can get latest news about the market as well blockchain technology.

However, most people are too busy read through the long articles which is a time-consuming process. Sometimes trader are also hard to grasp main bullet points of the long news. Such situation might put them at disadvantage position in crypto market.

What it does

This app retrieves the latest crypto news from CoinDesk by performing a keyword-based or URL-based search. Then, it creates downloadable vocal summaries of the news. So that trader can listen the vocal summaries in no time and act fast in their market trading activities.

Workflow

How I built it

I used python and Modzy SDK to build application backend logic and streamlit for its frontend GUI. Text-summarization and Text-to-Speech models are used to produce vocal summary. For serving, application is deployed to streamlit platform.

Challenges I ran into

Selenium usage to manipulate CoinDesk webpage
Chromium driver installation on remote streamlit instance during deployment

Accomplishments that I'm proud of

Integrated Modzy ML model to streamlit application and deployed it successfully
Found and reported bug to Modzy team to imrpove Modzy's ML models

What I learned

how to use streamlit to build and deploy machine learning application in short amount of time
how to use Modzy SDK as API to Modzy models and integrate them to streamlit frontend
selenium package for navigating and manipulating webpage elements