Novus AI - Enterprise Interview Suite

An adaptive, AI-driven interview simulator offering integrated DSA compilers, behavioral loop training, and live vocal tracking to replicate real-world technical assessments.

Comment

Inspiration

In a highly competitive and shifting tech job market, resilience is mandatory. Students and junior developers are under immense pressure to master data structures, system design, and behavioral interviews all at once. We wanted to build a unified, accessible workspace that replicates the intense environment of a real technical interview, helping candidates build the composure and skills needed to succeed without expensive coaching.

What it does

The Novus Interview Simulator is a comprehensive client-side interview workspace. It provides five dynamic preparation modes:

  • Algorithm Lab: Dynamic DSA problem generation.
  • Architecture Lab: System design frameworks and blueprints.
  • Communication Lab: Behavioral STAR method training.
  • Assessment Simulator: Replicated online assessments with a live, multi-language compiler.
  • Live Simulation: A full voice-to-voice interview room with real-time vocal stream tracking and client-side biometric attention tracking.

How we built it

We built the application prioritizing speed, accessibility, and privacy.

  • Frontend: Pure HTML, CSS, and vanilla JavaScript to keep the application lightweight and framework-agnostic.
  • Backend & Security: Supabase Edge Functions act as a secure proxy to handle database sessions and protect our API keys using Row Level Security (RLS).
  • AI & Logic: The Google Gemini API powers the dynamic generation of interview questions and evaluation rubrics.
  • Execution Environment: The Piston API enables the live code compilation within the browser.
  • Biometrics: TensorFlow.js runs facial expression inference entirely locally on the client's machine.
  • Analytics: Novus by Pendo is integrated to track user journeys, feature adoption, and session telemetry.
  • Deployment: Hosted live via GitHub Pages.

Challenges we ran into

Securing our API keys was a major hurdle. We initially faced CORS and security issues trying to call the Gemini API directly from the frontend. We overcame this by architecting a serverless backend using Supabase Edge Functions to safely proxy the requests. Additionally, configuring GitHub Pages to properly serve our relative file paths and bypassing Jekyll processing for our local machine learning models took significant debugging.

Accomplishments that we're proud of

We are incredibly proud of running local biometric facial-expression inference entirely in the browser. By keeping this client-side, we ensured low latency and high privacy for the user. We also successfully integrated a complex stack of AI generation, live code compilation, and enterprise-grade telemetry into a seamless, dark-mode UI using only vanilla web technologies.

What we learned

We deepened our understanding of serverless architectures, specifically how to secure client-heavy applications using edge functions. We also learned how to properly integrate and verify third-party enterprise analytics tools (Pendo) into a live production environment.

What's next for Novus Interview Simulator

We plan to expand the dynamic knowledge modules to include specialized tracks for DevOps and Cloud Architecture, and introduce historical trend graphs so users can track their algorithmic optimization speeds over time.

Built With

Share this project:

Updates