Pitch Perfect

Inspiration

The inspiration behind PitchPerfect came from the growing need for accessible, AI-driven interview preparation tools. Many job seekers struggle with anxiety, lack of feedback, or poor self-awareness during interviews. We wanted to create a platform that offers personalized, real-time feedback—so users can practice, improve, and land their dream roles.

Our goal was to build something that bridges the gap between casual preparation and professional coaching, using the power of AI and video analysis.

What it does

PitchPerfect is an AI-powered mock interview platform that allows users to record and submit responses in either:

Practice Mode: For casual self-improvement.
Interview Mode: For structured, simulated interview sessions.

After submission, users receive visual feedback on:

Sentiment of their answer
Relevance to the question
Confidence in delivery

The platform provides an interactive dashboard, scores performance, and helps users track growth over time.

How we built it

We brought together multiple open-source tools and models to build PitchPerfect:

Frontend/UI: Gradio for a clean and responsive interface
Transcription: Whisper (OpenAI’s ASR model)
Relevance Analysis: Ollama 3.2
Sentiment Analysis: VADER Sentiment

All of this is wrapped in a smooth, real-time processing pipeline to deliver immediate feedback to the user.

Challenges we ran into

Whisper transcription quirks: Handling variability in accent, pacing, and background noise sometimes impacted transcription accuracy.
Relevance scoring: It was challenging to find the right model that understands the semantic connection between a question and a response. We iterated multiple times before landing on a reliable setup using Ollama.

Accomplishments that we're proud of

Built a fully functional end-to-end AI mock interview platform in just one day
Seamlessly integrated multiple complex components (audio, video, NLP, UI)
Designed an experience that is useful, scalable, and easy to use for job seekers at any level

What we learned

The power of rapid prototyping using modular tools like Gradio
Fine-tuning user feedback systems to balance precision and usability
The importance of good model selection and prompt engineering for task-specific NLP analysis

What's next for PitchPerfect

We're just getting started. Here's what's coming up:

Interviewer mode: AI-generated follow-up questions and feedback from the interviewer's perspective
Question bank: Categorized by company, role, and difficulty
Progress tracking: Visual trends and insights over time
Custom feedback: Based on job role, industry, and seniority

Our vision is to make PitchPerfect a go-to platform not only for applicants but also for recruiters looking to streamline and enhance the hiring process.