MeetScribe

Inspiration

To prevent the spread of COVID-19, people are practicing self-quarantine while working from home remotely. This means all the meetings and interactions with colleagues are taking place online via video conferencing tools like Zoom, Hangouts, Microsoft Teams, etc. The information discussed during these meetings can be difficult to organize. Keeping up with post-meeting logistics like sending out meeting notes, summarizing action items, scheduling follow-up discussions and creating well-defined objectives is challenging. Hence, I created MeetScribe to help boost the productivity of professions working from home.

What it does

MeetScribe is a smart meeting assistant that provides a layer of intelligence during meetings and automates all the post-meeting tasks to drive higher productivity. It transcribes, summarizes and analyzes meeting notes in seven languages and provides an archive of meeting recordings. It performs all the tedious work during the meetings so they can actually be delightful for you.

How I built it

The technology stack used is as follows:

Render for hosting the web application
Angular4 + TypeScript to build the FrontEnd
Java to build the BackEnd
Google Could Speech-to-Text for Transcription

Challenges I ran into

Here are a few challenges that were encountered while developing this service:

Selecting a speech recognition engine with the highest accuracy.
Robustness around background noise and speaker with accents.
Accurate summarization of meeting notes.
Allows multiple attendees to join the same meeting and transcribe notes simultaneously.
Implement role-based access control for members.
Implementing the highest privacy and security standards.

Accomplishments that I'm proud of

Building a relevant solution that can have meaningful impactful for millions of professionals trying to flatten the curve is an inspiring pursuit. With this ambition in mind, I was able to build and deploy an end to end functional web application that completely platform agnostic and can be used on any device. I was able to build a lot of functionality in a condensed time frame. The entire architecture is quite modular and the service scales quite efficiently as multiple meetings are conducted.

What I learned

The best way to build an AI service is to use multiple microservices to get started and identify the core use cases for end-users that can deliver maximum impact.