Vode | Devpost

Recruiter Role View
Candidate View
Offerings
Home page
Recruiter Candidate View
Interview Mode

Inspiration

Every year, companies lose over 10,000 engineering hours conducting technical interviews, time that could otherwise go into innovation and product development. Even worse, the process is highly inconsistent: two candidates applying for the same role can face different interviewers, questions, and difficulty levels. This leads to unfair evaluations and unpredictable hiring outcomes.

We wanted to change that. We envisioned a platform that would standardize technical interviews without removing their human element — a system that could evaluate skill objectively, provide helpful feedback, and save recruiters and engineers hundreds of hours per hiring cycle. That idea became Vode.

What it does

Vode is an AI-powered interviewer built to conduct and assess technical interviews fairly, efficiently, and at scale.

It automates the full process from question selection to scoring and reporting while ensuring consistent difficulty across candidates. The AI engages conversationally, asks follow-up questions, and provides gentle hints when needed, simulating a real interview.

Recruiters can track candidate progress, engineers can configure question pools and scoring rubrics, and the system automatically flags borderline performances for human review. Every session includes code replays, transcripts, and structured analytics, making it easier to compare candidates across rounds.

How we built it

We used Django for the backend, cached imports for managing LLM memory, and Jinja2/Django Template Language, Bootstrap CSS, and VanillaJS for the front end.

For the AI, we integrated Gemini for question generation, reasoning, and code evaluation, Eleven Labs for realistic voice output, and internal logic to maintain fair question difficulty. This architecture allows Vode to handle concurrent interviews while providing a smooth, human-like experience.

Challenges we ran into

Designing database relations to track candidates, interviews, and question pools in real time.
Syncing the AI interviewer’s voice, reasoning, and scoring without latency.
Generating consistent metrics and analytics from cached transcripts and recorded sessions.
Calibrating the difficulty selection so no candidate gains an unfair advantage.

Accomplishments that we are proud of

Built a fully interactive AI interviewer that speaks, listens, and evaluates code in real time.
Achieved difficulty-consistent question selection for fairness across candidates.
Created a recruiter dashboard for live candidate tracking and engineer dashboards for reviewing flagged interviews.
Delivered a scalable system capable of running multiple interviews simultaneously with minimal resource overhead.

What we learned

We deepened our understanding of:

LLM prompting and conversational flow for real-time evaluations.
Caching mechanisms for scalable memory management.
Cloudflare API usage for storing blob data
The importance of human-in-the-loop fairness when building AI tools for recruitment.

What’s next for Vode

Next, we plan to expand beyond DSA interviews into:

System design interviews with whiteboard-style collaboration.
Pair programming simulations for team fit assessment.
Behavioral scoring add-ons powered by tone and reasoning analysis.

Our ultimate goal is to make Vode the standard AI interviewer for technical recruiting, saving time, ensuring fairness, and making technical hiring more transparent for everyone involved.