Shamith.ai

Shamith.ai is a personal interview coach! We used DeepGram & Gemini to make a coach that asks you resume questions, gives you a leetcode problems w/ hints & grading, & rates your interview ability.

Comment

Inspiration

Our project draws inspiration from our cracked friend Shamith. We are all terrible at leetcode and the original Shamith can't make child processes, so we made Shamith.ai instead.

What it does

Shamith.ai is a interview prepper with 3 phases:

  1. resume screening: first, Shamith reads your resume, and asks you using your speakers questions about your experiences. In between each question, he leaves time for you to answer, and notes your answer down for use later.

  2. technical round: Next, shamith gives you a leetcodd question and a text editor with some hidden test cases to run against. Every time you test, shamith will let you know how many tests you are passing with your speakers and if you are failing tests, he will give you hints to point you in the right direction. There's also a hint button so if you are really stuck, you can press the microphone and speak into the mic to ask Shamith a question and receive a response.

  3. rating finally, Shamith will let you know how your inverview went. He will rate you on a few metrics, how good your resume answers were, how good your code was, how many hints you asked for, etc...

How we built it

This entire project was a combinatino of DeepGram and Gemini being implemented on a React frontend with a Flask backend. Resume screening takes in a pdf and parses it to Gemini who reads it, gens questions, and reads them aloud with DeepGram. In between, deepgram listens. for you answers and takes notes of them for rating.

Technical test used a text editor library called CodeMirror and flask Express to run python code against our test cases. Gemini is queried when we hit test and during the hint button press. It is given the text editor content as context

the final rating uses all of our questions and answers and the hints we used to rate us using Gemini.

Challenges we ran into

Our biggest challenge was definitely figuring out how to use DeepGram's API. We kept running into all types of issues when implementing it with a flask backend.

Accomplishments that we're proud of

Our biggest most proud accomplishements are 100% when we completed our respective portions of the backend implementation and were able to speak wiht Gemini using deepgram.

What we learned

We definitely sharpened our team work and our ability to adapt to complext API's

What's next for Shamith.ai

SHamith.ai will continue to implement a difficulty slider, more comprehensive questions, and many more.

Share this project:

Updates