Inspiration

As international students, the biggest hurdle we face before coming to the US is the visa interview. As with any interview, you should not have a problem understanding the question, language, accent, etc. The best way to overcome it is to practice. Practice with people of different origins. We created this app to provide access to virtual agents of different origins. People hire a coach to help them practice for the interview and coach them through the job-search process. This app is a cost effective and a scalable solution.

What it does

The app provides a complete visa interview experience with a curated set of questions. The user can practice those sets of questions by selecting different faces, and the app using Deep Learning creates a video with that face to ask questions. Users can practice their answers by recording their audio and we provide sentiment analysis on that.

Receive feedback and tips on improving your interviewing skills so you'll be well equipped to handle an actual interview.

How we built it

The web app uses Flask for backend connections. The videos are created using Deep Learning based on [1]. PyTorch deep learning libraries, numpy, ffmpeg, etc create the videos. The only required input for creating a video is a base video and an image of the new face. Speech recognition software converts the audio into text.

Challenges we ran into

These are the challenges we faced: Create a set of questions. For this, we researched questions commonly asked in Visa Interviews. Create video samples to cover people of different origins. Using Deep Learning, we converted images to real-life videos to ask interview questions. Web app integrations had challenges to make different software work together.

Accomplishments that we're proud of

We are proud of the fact that we have created a solution to a problem which we faced. Visa interviews are stressful for some people, and we think practicing on our platform can provide valuable feedback.

What we learned

We got to use our deep learning skills in an application. We learned ways to integrate different libraries and mini-apps into one.

What's next for Interview Prep

Next is to get user feedback to improve the app. We would also like to add a feature where user can upload an image and we create the video instantly. Another feature will be to record video responses and analyze facial expressions.

References

  1. https://aliaksandrsiarohin.github.io/first-order-model-website/
  2. https://github.com/app-generator/flask-material-dashboard
  3. https://cloud.ibm.com/apidocs/tone-analyzer?_ga=2.160634813.1437600724.1618736217-2131175079.1618736217&code=python
Share this project:

Updates