DubHacks 2025 🚀

This is our hackathon project for DubHacks 2025.

Team

• Trisha Bhatawdekar
• Neha Dubhashi
• Misha Nivota

Tech Stack

• [e.g. React, Node.js, Python, etc.]

Core system

• The user sits (via webcam + mic) and conducts a mock interview — either a pre‑set question bank (e.g., “Tell me about a time you…”, “Why do you want this role?”) or an open prompt.

• Using audio ML you analyze: tone of voice, filler words (“um”, “uh”), pacing, volume variation, clarity of speech, confidence cues, maybe signs of nervousness (pauses, hesitation).

• Using visual ML you analyze: face (eye contact, head nodding/shaking, micro‑expressions, smiling), body posture (leaning in/out), gesture frequency, facial orientation (looking at camera vs away), blink rate maybe.

• Combine these into descriptive feedback: e.g., “You used filler words 12 times in 90 seconds”, “You maintained eye contact ~70% of the time”, “Your voice volume dipped significantly in the last answer”, “Your posture is leaning back which may signal disengagement” etc.

• Then provide actionable suggestions: “Try to reduce filler words by pausing instead of “um” when thinking”, “Lean in slightly and keep shoulders square to camera”, “Increase vocal variety: you sounded quite flat through the middle”.

• Optionally, score the interview according to multiple axes: Confidence, Clarity, Conciseness, Engagement, Body Language. Provide benchmark/ranking vs peers.

• Provide history/dashboard: track user progress over multiple sessions, show improvements (e.g., filler words per minute dropped from 10 to 6, average eye contact from 50% to 75%, etc.).

• Offer variant modes: e.g., behavioral interview, technical interview (where you ask coding questions and they respond), role‑play with a virtual interviewer (simulate tough scenario).

• Optionally integrate language or accent support (help non‑native speakers improve clarity), or industry/domain specific (sales interviews vs engineering vs leadership).

• UI/UX design: simple webcam + mic record, playback with annotated timeline (show when filler word occurred, when gaze drifted, when posture changed). Provide transcripts of responses, highlight key phrases (maybe “impact‑oriented phrasing”, “STAR technique” compliance).

Why it’s good / what problem it solves

• Many people are nervous about interviews and don’t have affordable or high‑quality coaching. A semi‑automated coach lowers barrier.

• Visual + audio feedback is hard to self‑get; recording yourself is okay but you may not catch your own body language or voice patterns.

• Helps quantify improvement and motivate practice (gamification, progress tracking).

• Could be used by students, job‑seekers, people switching careers, non‑native speakers.

Troubleshooting

• npm run dev freezes at Starting...
  • clear the npm cache (Remove-Item -Recurse -Force .next)