Introduction to
Dibyanshu Chatterjee

Introduction to Dibyanshu ChatterjeeIntroduction to Dibyanshu ChatterjeeIntroduction to Dibyanshu Chatterjee

Introduction to
Dibyanshu Chatterjee

Introduction to Dibyanshu ChatterjeeIntroduction to Dibyanshu ChatterjeeIntroduction to Dibyanshu Chatterjee
  • Home
  • Projects
  • Research
  • Contact Me
  • More
    • Home
    • Projects
    • Research
    • Contact Me
  • Home
  • Projects
  • Research
  • Contact Me

try the app

Click Here

Building an Interview Bot with OpenAI GPT-3.5 and Streamlit

In the era of AI, I embarked on a journey to create an interview bot that leverages the power of OpenAI’s GPT-3.5 model. The goal was to design a system that could generate dynamic interview questions based on a user’s resume, job dehttps://websites.godaddy.com/interviewbotscription, company name, and interview type. The finished app is hosted on streamlit at: https://interview-bot-verbal-nonverbal-eam3i9niz8jcqnqjnwrubg.streamlit.app/. Here’s how I did it.

The Central Brain: GPT-3.5

Designed a series of posters for a music festival featuring a unique illustration style and bold typography. The posters were used for both print and digital promotion.

Indexing Over Custom Data: Llama Index

The Llama Index was used for indexing over custom data. This powerful tool allows for efficient searching and retrieval of information from the custom data, enhancing the bot’s ability to generate relevant questions.

The Interface: Streamlit

Streamlit, a fast and easy way to create apps, was used for constructing the interface of the bot. The interface asks the user to upload their resume, job description of the interview, the company name, and the type of interview (Technical, Behavioural, etc.). It also provides an option to practice for “verbal” interviews or non-verbal/technical ones.

Building Custom Context

The bot builds a custom context based on the user’s resume, the job description and other details filled at the homepage of the app. This context is then fed into the GPT-3.5 model to generate the interview questions. The context is built using the PdfReader from the PyPDF2 library to extract text from the user’s resume and the text from other inputs are combined. This is done to ensure agility of question generation and answer evaluation.

Converting Questions to Audio

To make the bot more interactive and user-friendly, I incorporated a feature to convert the generated questions into audio. This is done particularly for the verbal module, where the user can listen to the audio and record their voice response. This was achieved using the Google Text-to-Speech (gTTS) library, which converts the text into speech audio.

Checking OpenAI API Key

The bot checks the validity of the OpenAI API key before retrieving the GPT-3.5 model. This is an essential step to ensure that the bot can successfully connect to the model and generate the interview questions.

Checkout the code on GitHub

code

Powered by

This website uses cookies.

We use cookies to analyze website traffic and optimize your website experience. By accepting our use of cookies, your data will be aggregated with all other user data.

Accept