A Windows desktop application for realistic, AI-powered interview practice — built with Electron and Python.
This repository is private and the source code is not publicly available.
This README is published for portfolio and demonstration purposes only. The codebase is kept private to protect the implementation. If you are a recruiter, collaborator, or evaluator and would like to review the source code, please reach out directly.
This project was developed as a final-year college project at the Department of Computer Science & Engineering, Govindrao Wanjari College of Engineering and Technology, Nagpur (affiliated with RTM Nagpur University).
The research paper based on this project — "AI-Powered Mock Interview Assistant" — was submitted, reviewed, and presented at ICRTSET-2026 (International Conference on Recent Trends in Science, Engineering and Technology) on April 6, 2026, and has been accepted for publication.
| Detail | Info |
|---|---|
| Institution | Govindrao Wanjari College of Engineering & Technology, Nagpur |
| Department | Computer Science & Engineering |
| University | RTM Nagpur University |
| Conference | ICRTSET-2026 |
| Presented | April 6, 2026 |
| Paper Status | Accepted for Publication |
AI Mock Interview Assistant simulates a complete guided interview session on your desktop. Select topics, set a difficulty level, answer questions verbally or through the built-in code editor, and receive a structured PDF performance report at the end of every session.
The application follows a local-first architecture — there is no application-owned cloud backend. All user data stays on your machine.
- UI runs inside Electron
- AI engine runs in Python
- All user data and reports stored locally
- AI requests routed through your own Groq API key
- Topic-based session configuration from a central dashboard
- Domains: Programming, CS Fundamentals, Frameworks, SQL, Behavioral, Leadership
- Difficulty control before session start
- One-question-at-a-time flow with adaptive follow-up questions
- Built-in code editor for coding-style prompts
- Live speech-to-text transcription via Groq Whisper
- LLM-powered question generation and answer evaluation
- Text-to-speech playback for interviewer questions
- Real-time microphone activity visualization
- Fullscreen interview mode
- Start, pause, resume, and end controls with a live timer
- Camera preview in dashboard and session views
- Camera and microphone device selection from Settings
- Focus warnings triggered when the window loses focus
- Camera-based face-presence detection during sessions
- Filler-word tracking across all answers
- Behavioural flag summary included in the final report
- Auto-generated PDF report on session completion
- Per-question scores and structured feedback
- Overall performance summary with recommended next steps
- Behavioural analysis section covering focus warnings and filler-word usage
| First-Time Setup | Dashboard |
|---|---|
 |
 |
| API key onboarding on first launch | Topic & difficulty selection dashboard |
| Interview Session | Settings |
|---|---|
 |
 |
| Fullscreen interview mode with live timer | Camera & microphone device selection |
| PDF Report |
|---|
 |
| Auto-generated performance report with per-question scores and behavioural analysis |
flowchart TB
subgraph electron["🖥️ Electron Desktop Layer"]
direction TB
UI["Dashboard & Onboarding UI\n(index.html · renderer.js)"]
WM["Window Manager\n(main.js)"]
DEV["Device Selection\nCamera · Microphone"]
IPC["IPC Bridge\nstdin / stdout"]
WM --> UI
UI --> DEV
UI --> IPC
end
subgraph python["🐍 Python AI Engine (engine/app.py)"]
direction TB
SM["Session State Manager"]
QG["Question Generator\nLLaMA 3 via Groq"]
STT["Speech-to-Text\nGroq Whisper"]
FU["Follow-up Logic\nContext-aware"]
MON["Behaviour Monitor\nFocus · Face · Fillers"]
PDF["PDF Report Generator\nReportLab · Pillow"]
SM --> QG
SM --> STT
QG --> FU
STT --> FU
SM --> MON
SM --> PDF
end
subgraph external["☁️ External"]
GROQ["Groq API\nWhisper + LLaMA 3"]
end
subgraph local["💾 Local Storage (AppData)"]
CFG["config.json\nAPI Key"]
USR["user.json\nProfile"]
RPT["reports/\nPDF Files"]
end
IPC <-->|"Process messaging"| SM
QG & STT -->|"API requests"| GROQ
SM --> CFG & USR & RPT
| Layer | Technology |
|---|---|
| Desktop UI | Electron, HTML, CSS, JavaScript |
| AI Engine | Python, Groq SDK, NumPy, sounddevice |
| Speech-to-Text | Groq Whisper API |
| Report Generation | ReportLab, Pillow |
| Configuration | python-dotenv |
| Packaging | PyInstaller, electron-builder |
.
├── assets/
│ ├── icon.ico
│ └── style.css
├── build/
│ └── installer.nsh
├── engine/
│ ├── app.py
│ ├── engine.spec
│ └── runtime_hook_ssl.py
├── index.html
├── main.js
├── renderer.js
├── requirements.txt
├── package.json
├── build.bat
└── LICENSE.txt
First Launch
└── Check for saved Groq API key
├── No key → Onboarding / setup screen
└── Key found → Load user profile → Dashboard
Session
└── Select topics & difficulty
└── Fullscreen interview mode
├── Engine asks one question at a time
├── User answers verbally or via code editor
├── Engine transcribes + generates follow-ups
└── Focus & behaviour signals tracked
End of Session
└── Engine finalises Q&A set
└── Evaluate answers
└── Generate & open PDF report automatically
On first launch, the application prompts for a Groq API key. This is by design — the app does not ship with an embedded key, as doing so in a desktop binary is a security risk. Each user authenticates AI usage through their own Groq account.
Keys begin with gsk_ and are stored locally at:
%AppData%\AI Mock Interview Assistant\config.json
| File | Path |
|---|---|
| API key | %AppData%\AI Mock Interview Assistant\config.json |
| User profile | %AppData%\AI Mock Interview Assistant\user.json |
| PDF reports | %AppData%\AI Mock Interview Assistant\reports\ |
| Engine crash log | %AppData%\AI Mock Interview Assistant\engine-crash.log |
| Engine stderr log | %AppData%\AI Mock Interview Assistant\engine-stderr.log |
Privacy: All data — API key, user profile, and session reports — is stored locally on your machine. No data is sent to any application-owned server.
Reports are generated automatically at the end of every session and opened immediately. Each report contains:
- Candidate name, interview topics, and difficulty level
- Question count, filler-word count, and focus warning count
- Overall rating and per-question scores
- Structured feedback and improvement tips
- Behavioural analysis section
- Recommended next steps
Font support: The PDF engine uses Product Sans when engine/fonts/ProductSans-Regular.ttf and engine/fonts/ProductSans-Bold.ttf are present, and falls back to Helvetica automatically if they are not.
pyinstaller is not recognized
Use build.bat. The script installs and runs PyInstaller from the project virtual environment directly.
401 invalid_api_key error
Verify the saved key is a valid Groq key beginning with gsk_. Update it from the in-app Settings screen.
Session does not start
Confirm at least one topic is selected and the Python engine is running. Rebuild with build.bat if the engine bundle is outdated.
PDF report is not generated Check the log files at:
%AppData%\AI Mock Interview Assistant\engine-crash.log
%AppData%\AI Mock Interview Assistant\engine-stderr.log
- Demo video walkthrough (full session + PDF report)
- Research paper publication link
- Screenshot-rich README
- Refined onboarding and first-run UX
- Expanded interview packs with curated content
- Stronger report visualisation and analytics
- Optional managed backend mode
- Session history export and import
This project is proprietary and not open source. The source code is kept private.
See LICENSE for full legal terms. In summary:
- Personal and educational reference use only
- Redistribution is not permitted without explicit written permission
- Commercial use is not permitted without explicit written permission
- Modification and derivative distribution are not permitted without explicit written permission
Ved Sharanagate
B.Tech, Computer Science & Engineering
Govindrao Wanjari College of Engineering and Technology, Nagpur
📧 vedsharangate05@gmail.com