Scaling Potato (Interview Prep with AI)

Inspiration

While applying to software engineering jobs and preparing for technical interviews throughout our (young) career, we found that platforms like Leetcode didn't fully prepare us for a real technical interview.

We created Scaling Potato as a modern solution to help people prepare for technical interviews by practicing with our AI interviewer.

What it does

Users can practice a mock technical interview with our AI-guided interviewer. We try to make this as seamless as possible, allowing the user and the interviewer to converse and solve a coding problem. The user can also use the built-in code editor we added to write and execute their code for the problem. We crafted strategic questions asked in real interviews and at the end, our platform analyzes the entire practice interview data to crunch out actionable results and give feedback.

✨ After using our product, we hope that job seekers gain more confidence in their interviews, improve their technical communication skills, and problem solving ability.

How we built it

We created a NextJS application deployed on Vercel.
We synthesize text-to-speech in real time using Google's text-to-speech API.
We analyze speech-to-text in real time using AssemblyAI's streaming API.
We used GPT-3 with thoughtfully curated prompts to analyze interview results.
We used Monaco Editor and Piston to enable users to write code and execute it in their browser instantly.

Challenges we ran into

Making our product feel interactive and conversational for the user while conversing with the AI. We wanted to simulate talking to a real person the best we can with the constraints that we had. So we placed a huge focus on subtle UI features and personalization like a minimal UI, subtle animations, auto-typing text, text-to-speech, and speech-to-text.
The audio (speech-to-text and text-to-speech) was really tricky, we ran into some problems using it in NextJS but figured it out eventually.
Managing feature creep.

Accomplishments that we're proud of

We're proud of a lot! Mainly using so many APIs in a real-world use case in such a limited time.
Our UI design was clean and minimal, we crafted something we are proud of in an MVP product.

What we learned

We learned a bunch about new APIs (AssemblyAI, Google, GPT-3, Piston).
Doing a lot of user experience research to find what works for our target audience.