Technical interview helper

#idea

• Listen to conversation and answer technical questions
• Train MOE model
• You can train a model to classify a question to a leetcode question (embed the type of problem, ie sliding window, two pointer, etc)
• Use NVIDIA Eye Contact system to look like you’re looking at the screen https://developer.nvidia.com/blog/improve-human-connection-in-video-conferences-with-nvidia-maxine-eye-contact/
• Not let it be seen by screen sharing

Justification for why to build this

• Current hiring systems do work, but they don’t get the whole story across. Along with pure intelligence are things like tenacity, vision, etc that can’t be tested for enough in interviews.
• The truth is, I do think most interviews are as good of a judge of ability/character you can get in a short amount of time. The correct thing to do is fire fast afterwards.
• So then why build a tool that will get less skilled people into jobs that require high skill?
• One reason could be that others are going to be doing this as well. It would be nice to beat them, own the market, and the figure out how to solve this issue from the inside.
• Another way to look at it is that it will get people employed. Many people are unemployed and many job openings exist. Thus, it isn’t a zero sum game right now. The bar is too high and needs to meet the market.
• I can envision an interesting future where we get lots of people using it at company X, then approach company X saying this is what’s happening, but we can solve this with our own AI interviewer—essentially giving them an option to get the best candidates again.
  • Or maybe they have to pay some fee for AI protection, and our product + all the others won’t work on it
• The one part where AI will suffer is generating personality. If we look towards a future where most labour (digital or physical) will be automated, the true test of whether to work with someone will be that of their personality. If anything, this idea is building infrastructure for this future world

Why me?

• Waterloo student = tons of experience in the job search field + tons of access to customers
• This most likely isn’t my magnum opus, but as I want to practice growth and this doesn’t seem like a crazy technical problem, it should be fun to work on while in school

Features

Avoid screenshare issues with: https://github.com/Stengo/DeskPad

• Interview Copilot
  • Connect to a call
    • Some ways to go about this
      • Extension that hooks into websites like coderpad, hackerrank, etc. Since it’s a part of the webpage, it won’t be able to tell if you’ve left the page.
        • Concerns
          • What if we don’t support the platform (vscode, google colab)
            • even for non-technical interviews
          • Share screen
          • Custom UI for every app might be cumbersome
            • especially if these UI changes
      • Desktop app that’s an overlay on a regular google meet call + coding interview
        • Concerns
          • Window management
            • if you ever need to click into our UI, will the coding platforms detect that
          • Share screen
            • Maybe we can get around this
          • Screenshots to get text could pose difficult
            • occlusion
            • poor OCR
      • Finalround type: website in separate tab, share the google meet with product, then picture in picture support
        • Concerns
          • Share entire screen (window level doesn’t show)
    • Need to support:
      • Zoom
      • Google Meets
      • Microsoft
    • Coding platforms
      • Hackerrank
      • Coderpad
      • Leetcode
      • Replit
      • codeshare
      • google docs
  • Eye gaze redirection
• Preparation
  • Input resume for better behavioural answers
  • Find problems seen from that company before
• Application
  • AI Generated resume
  • reorganize depending on job
  • add keywords

Focus on technical interviews first: more of a moat, more tangible results, typically the tech community disperses products to regular people (chatgpt).

What are some common things our friends do during interview season:

1. Go on to reddit and and discord groups to get updates on potential kinds of questions + get connected to people who have recently interviewed - get hints on what kind of questions to prepare
2. Ask friends to comment on different kinds of resumes, or give information on what kind of resumes to build for what roles.
3. Build projects/research things specific to a role
4. Create multiple resumes and cover letters based on job descriptions→ use information about tickets solved to update resume, talk to gpt and ask it to generate an optimal experience point, then refine it, and get feedback on resume
5. Grind leetcode and neetcode (get leetcode premium to find company specific problems)
6. Find resources for systems design interviews
7. Research companies and reach out to multiple people on LinkedIn - draft cold intro messages specific to them
8. Connect with mutuals, or alums/alma mater on LinkedIn to talk about upcoming interviewers, and ask for referral.
9. Look for and reach out to job postings on X, Linkedin, and externally, and manage a spreadsheet of all this information in one place.
10. Ask for tips from others who do well, on how best to interview, potentially do mock interviews based on tips.