codeverse

Inspiration

Technical interviews are notoriously difficult—not just because of the coding challenges, but because candidates must also clearly articulate their thought process under pressure. While platforms like LeetCode help with problem-solving, they don’t train candidates to communicate their reasoning in real time. Behavioral interview mock platforms exist, but there’s no equivalent for technical interviews that combines live problem-solving with structured AI-driven feedback.

Codeverse was inspired by this gap. We wanted to create an interactive, AI-powered experience that not only helps candidates solve coding problems but also strengthens their ability to think out loud and communicate well, a vital skill for interviews. By simulating a live technical interview, Codeverse helps candidates refine their approach, build confidence, and perform better in high-stakes interviews.

What It Does

Codeverse is an AI-powered technical interviewer designed to simulate software engineering and financial interview scenarios. Users receive a technical problem and, as they work through the challenge, the platform listens to their thought process via speech-to-text, ingests their code, and then uses a Gemini API-backed engine to facilitate real-time audio communication, just like in a live interview! The AI responds dynamically, asking follow-up and clarification questions throughout, giving hints when needed, and providing feedback at the end of the interview, helping users refine both their technical and verbal skills in a realistic interview setting and making the interview preparation process more interactive and accessible.

How We Built It

To manage the scope, we divided the project into clear modules: problem presentation, real-time code and audio processing, and feedback generation. We built our frontend, including a code editor with multi-language syntax highlighting, with TypeScript aided by Mantine libraries. Our Flask backend leverages real-time text-to-speech with multithreading, the Gemini API for AI-driven responses, and the Google Cloud Text-To-Speech AI for fast, responsive, and human-like conversation. Finally, we connected everything with Socket.io for rapid, bi-directional communication.

Key Features & Challenges

Key features include the live interview simulation, real-time code evaluation, and a dynamic feedback loop that adapts to both audio and code inputs. One of our biggest challenges was ensuring seamless integration between diverse technologies, synchronizing audio feedback with code analysis in real time to accurately simulate the live interview experience with back-and-forth conversations. Through iterative testing and a collaborative approach, we refined our system to deliver an experience that mirrors a live technical interview.

Next Steps

We plan to expand codeverse by incorporating personalized learning paths, additional problem categories such as quant, data science, and cybersecurity, and enhanced analytics that track progress over time. Our ultimate goal is to make technical interview preparation not only effective but also engaging, empowering users to achieve their career aspirations with confidence.