LowkeyPrepped

With a projected shortage of 85.2M software engineers by 2030, we built LowkeyPrepped to provide AI-assisted, realistic mock interviews, helping aspiring engineers land their dream job.

Comment

Inspiration

It’s 2030: The tech industry is booming, but there’s a catch—there’s a global shortage of 85.2 million software engineers. Companies are struggling to fill critical roles, and aspiring engineers, despite their passion and potential, face daunting interviews that could make or break their chances of landing their dream job. Nerves, inconsistency, and lack of preparation are holding many back. The demand is there, but the path to success is filled with challenges.

What it does

LowkeyPrepped redefines interview preparation with an AI-powered mock interview platform designed specifically for software engineers. It’s not just another practice tool; it’s a personalized coaching experience that delivers realistic mock interviews covering both technical and behavioral aspects.

Our conversational AI simulates real-world interview scenarios, providing instant feedback on responses, highlighting areas for improvement, and even offering guidance when a user gets stuck. Technical interviews are challenging, but LowkeyPrepped breaks them down, helping users tackle everything from coding questions to system design.

How we built it

To visualize the flow between different software systems, we developed a detailed systems design.

LowkeyPrepped Tech Stack Flow Chart

Our comprehensive tech stack:

Frontend:

  • Next.js with React for a responsive and dynamic user interface
  • TailwindCSS for efficient, customizable styling
  • Framer Motion for smooth animations

Backend:

  • Convex for database management and backend logic
  • OpenAI GPT-4o mini for conversational AI and interview question generation
  • Windows Speech Recognition for speech-to-text transcription during interviews
  • Google Text-to-Speech API for converting responses into audible feedback

Challenges we ran into

  • Learning Convex and integrating it into our project as the backend; while it simplifies certain aspects, it still posed a steep learning curve for a hackathon
  • Facing deployment issues, including linting errors and differences between production and development environments
  • Building a low-latency conversational tool was challenging, as creating the pipeline from speech-to-text, to an LLM, and back to text-to-speech required significant trial and error

Accomplishments that we're proud of

  • Learning Convex in just 4 hours and figuring out how to integrate it into our project
  • Creating a way for the AI assistant to display and format code properly, ensuring it doesn’t read the code aloud to make the interaction feel more natural and human-like
  • Integrating a user authentication system using Clerk and Convex, along with a well-designed UI and a proper routing system

Convex Challenge Submission

Use of Convex

  • We connected Convex as a webhook to Clerk, allowing users to sign in through Clerk and create entries in the user tables
  • Each user can create multiple interviews, with every message stored as a row in the interview table
  • We use three main tables—users, interviews, and messages—to organize data efficiently
  • This setup simplifies our backend system, making it easy to manage using TypeScript

Innovation

Tackles the growing demand for more immersive and effective interview preparation:

  • Introduces an AI interview assistant that simulates real-life interview dynamics
  • Transforms technical assessments with live coding and real-time feedback
  • Implements AI-driven hints to guide users through challenging questions

Technical Complexity

  • Fully utilized Convex mutations to update and manage interview and message data, queries to retrieve user-specific interview records, and actions to handle asynchronous processes efficiently
  • Implemented a 1:N relationship between Users and Interviews, and Interviews and Messages, using three key tables for the project
  • Each interview stores essential information such as difficulty, type (e.g., behavioral or technical), and programming language (for technical interviews)
  • Interviews are linked to users through a foreign key in the interview table, and messages are linked to their respective interviews via a foreign key in the message table

Design and User Experience

Our design emphasizes ease of use and efficiency:

  • Developed a clean, intuitive interface optimized for seamless user interaction
  • Implemented both light and dark mode options for user preference
  • Designed a UI for technical interviews that closely mirrors real-world interview environments, providing a familiar and practical experience

Impact

LowkeyPrepped addresses a critical need in the tech industry:

  • Targets the 80% of interviewees who struggle with inconsistency across interviews, aiming to improve confidence and reliability in the process
  • Aims to prevent the U.S. Labor Department’s projected global shortage of 85.2 million software engineers by 2030
  • Empowers aspiring software engineers to overcome interview anxiety and secure the roles the world needs filled

What we learned

  • It's challenging, but possible, to quickly learn and implement a brand new tech stack
  • AI has proven to be an effective interviewer, with the potential to improve further through refinements
  • Transitioning from development to production can be a complex and intricate process

What’s next for LowkeyPrepped

  • Partner with tech companies to offer first-round interviews directly through the platform
  • Implement a detailed transcript and highlight report feature that provides users with a comprehensive summary of their interview
  • Introduce live peer-to-peer mock interviews with instant AI feedback

Built With

Share this project:

Updates