InterviewSTAR

Inspiration

One of the most important professional soft skill is the ability to speak fluently and confidently about your experiences. Having this ability can open numerous doors, especially for recent college graduates or new entrants to the job market. This soft skill is especially important now given the current hiring climate. To gain an edge, why not use the current wave of easily accessible AI tools to your advantage? Why not have interview practice sessions on demand? With feedback directly actionable feedback too?

What it does

InterviewSTAR provides a platform to practice common interview techniques such as the S.T.A.R method or answering case studies. With enough practice, you can become more comfortable speaking about your experiences in a structured manner. From this level of comfort, you can derive confidence. Confidence that will show in your next interview or feedback session with your manager (did somebody say raiseee 💰)

How we built it

I used the Voiceflow API to get access to an LLM then I directly integrated a popular LLM in order to get faster responses. For the web app, I used Next.js with Tailwind + DaisyUI. For speech recognition and Text-to-speech, I used a mixture of GCP, Azure services and the OpenAI Whisper model.

Challenges we ran into

One of the hardest parts was synchronizing the state of all the components in the project. Another challenge was designing the prompts to provide the required depth of questions and feedback without exceeding the token limit or being overly verbose.

Accomplishments that we're proud of

Despite being a late submission, I'm super proud of seeing the project through and refining an idea I had in mind for a while.

What we learned

I learnt a lot about prompt engineering, specifically refining prompts and tweaking the params (top k, temperature etc) in order to get the required amount of details in an answer. I also learned more about including guardrails in prompts to reduce the likelihood of bad actors to cause havoc.

What's next for InterviewSTAR

Some impending updates will see :

• more interview methods being built out and introduced.
• better feedback with quantifiable steps for improvement and 🥁🥁🥁🥁 high scores!
• the ability to save and review practice sessions
• the ability to specify the difficulty level of the session or the tone of the speaker
  • multiplayer practice sessions

Built With

• azure
• gcp
• python
• react
• voiceflow