🎙️ Interview Agent AI — Intelligent Mock Interview & Candidate Evaluation Platform

Transform technical and behavioral interview preparation with autonomous AI assessment, real-time speech-to-text dictation, structured rubric scoring, and instant panel hiring recommendations.

🚀 Live Production App: https://interview-agent-ai.netlify.app/ | 💻 GitHub Source: https://github.com/bikram73/Interview_Agent_AI

📑 Table of Contents

📖 Description	🚀 Section
View the project features and capabilities. 👉
View the technologies, frameworks, and programming languages used. 👉
Explore the project's folder and file organization. 👉
Follow the installation steps and local development setup. 👉
Understand the complete AI document processing pipeline. 👉
Learn about the AI prompting strategy and anti-hallucination techniques. 👉
Understand how confidence scores are calculated and interpreted. 👉
View all deliverables required for the AI challenge. 👉
See an example of the structured JSON output generated by the AI agent. 👉
View the available REST API endpoints and usage examples. 👉
View an overview of the sample documents and extraction results included in this project. 👉
Open the dedicated Sample Results page to view all sample documents, images, PDFs, and extracted JSON outputs. 👉
Explore the complete system architecture, AI workflow, processing pipeline, data flow, deployment design, and technical decisions. 👉
Review implementation details, AI pipeline, performance metrics, benchmarking, validation strategy, privacy, testing, and technical specifications. 👉
Review processing speed, latency, and performance benchmarks. 👉
Understand the current limitations and known failure cases of the AI extractor. 👉
View the project license information. 👉

✨ Project Features & Capabilities

🎯 Multi-Track Role Library: Over 11+ curated technical and soft-skill tracks (Python Developer, Java/Spring Boot, Frontend/React, Backend APIs, Full Stack, AI Engineer, ML Engineer, Data Scientist, Cybersecurity Analyst, Cloud DevOps, HR Behavioral) plus support for Custom Job Roles.
📊 Dynamic Experience Calibration: Configurable difficulty curves tailored for Fresher / Entry Level, Junior (1-2 yrs), Mid-Level (3-5 yrs), and Senior (6+ yrs).
⚡ Structured AI Question Generation: Leverages Google Gemini 2.5/3 Flash with strict schema constraints to generate balanced interview sets spanning fundamentals, system design, architectural trade-offs, and STAR scenarios.
🎙️ Voice Dictation (Speech-to-Text): Built-in Web Speech API recognition enabling candidates to speak naturally in real-time, just like in a live interview room.
⏱️ Real-Time Timed Sessions: Built-in visual countdown timer, live word/character counters, and question sequence trackers.
🔍 Instant Question-by-Question Evaluation: Deep rubric grading with numerical scores (0–10), key strengths identified, missing technical gaps highlighted, and actionable STAR improvement tips.
💡 Ideal Response Benchmarking: Displays key technical concepts and ideal bullet points expected by senior interviewers.
📈 Executive Hiring Report: Synthesizes session transcript into an aggregate candidate score (0–100%), hiring recommendation (Strong Hire, Hire - Low Risk, Borderline, Needs Development), confidence level, and prioritized study roadmaps.
📥 Themed PDF & Text Export: One-click download of high-resolution PDF evaluation dossiers styled in the exact visual theme of the platform (score cards, gauges, strengths, weaknesses, and full transcripts) alongside plain .txt format.
🚀 Dual Deployment Ready: Seamless execution across Express.js servers (Docker/Cloud Run) and Serverless Netlify Functions with automated redirects.

🛠️ Tech Stack & Languages

+-------------------------------------------------------------------------+
|                              FRONTEND LAYER                             |
|  React 19.0.1  *  TypeScript 5.8  *  Tailwind CSS v4  *  Lucide / Icons |
|  Motion (Framer)  *  Web Speech API  *  Vite 6 Bundler                  |
+-------------------------------------------------------------------------+
                                    |
                                    v
+-------------------------------------------------------------------------+
|                             API & SERVER LAYER                          |
|  Node.js (v20+)  *  Express 4.21  *  Netlify Serverless Functions       |
|  @google/genai TypeScript SDK  *  esbuild Bundler  *  tsx Runtime       |
+-------------------------------------------------------------------------+
                                    |
                                    v
+-------------------------------------------------------------------------+
|                                AI CORE LAYER                            |
|  Google Gemini 3.6 / 2.5 Flash  *  Structured JSON Schema Enforcement   |
|  Prompt Engineering & Anti-Hallucination Grounding                      |
+-------------------------------------------------------------------------+

💻 Programming Languages

TypeScript (v5.8.2): 100% strict type safety across client UI, data models, server routes, and Netlify Functions.
JavaScript (ES Modules & CJS): Modern ES2024 runtime specifications.
HTML5 & CSS3: Modern semantic markup styled with Tailwind CSS v4 CSS design tokens.

📦 Key Libraries & Frameworks

Layer	Technology	Purpose
Client UI	`React 19.0.1`	Component-based reactive user interface
Styling	`Tailwind CSS 4.1.14`	High-performance CSS utility styling engine
Animation	`motion 12.23.24`	Smooth screen and state transitions
Icons	`lucide-react` & Google Material Symbols	Clean, professional UI icons
AI SDK	`@google/genai 2.4.0`	Official Google Gemini GenAI SDK
Dev Server	`Vite 6.2.3` & `tsx 4.21.0`	Ultra-fast HMR and TypeScript execution
Backend Server	`Express 4.21.2`	RESTful API proxy and static asset delivery
Serverless	`@netlify/functions 5.3.0`	Serverless cloud functions for Netlify hosting

📂 File Structure

interview-agent-ai/
├── netlify/
│   └── functions/
│       ├── evaluate-answer.ts      # Netlify function for candidate answer evaluation
│       ├── generate-questions.ts   # Netlify function for AI question generation
│       └── generate-report.ts      # Netlify function for executive summary & report
├── public/
│   ├── _redirects                  # Netlify SPA and API rewrite rules
│   └── vite.svg
├── src/
│   ├── components/
│   │   ├── EvaluationView.tsx      # Step-by-step evaluation, scores, and strength breakdown
│   │   ├── LandingView.tsx         # Hero landing view with role tracks & stats
│   │   ├── Navbar.tsx              # Application header navigation
│   │   ├── PerformanceView.tsx     # Final candidate report, hiring recommendation & transcript
│   │   ├── RolesView.tsx           # Role selector, experience configuration & question count
│   │   └── SessionView.tsx         # Active interview room, voice STT & timer
│   ├── services/
│   │   └── api.ts                  # Client-side API client wrapper with fallback resilience
│   ├── App.tsx                     # Main state machine, screen routing & session coordination
│   ├── index.css                   # Tailwind CSS v4 entrypoint & theme variables
│   ├── main.tsx                    # React DOM root entry
│   └── types.ts                    # Global TypeScript interfaces & data contracts
├── .env.example                    # Environment variable template
├── .gitignore                      # Git exclusion rules
├── ARCHITECTURE.md                 # System architecture, data flow & deployment docs
├── LICENSE                         # MIT open-source license
├── metadata.json                   # Applet configuration & server permissions
├── netlify.toml                    # Netlify build, redirects, and function routing
├── package.json                    # Dependencies, scripts, and build configuration
├── README.md                       # Comprehensive project documentation
├── SAMPLE_RESULTS.md               # Sample questions, candidate answers & evaluations
├── server.ts                       # Express.js full-stack development & production server
├── TECHNICAL_REPORT.md             # Benchmark metrics, AI pipeline & evaluation analysis
├── tsconfig.json                   # TypeScript compiler configuration
└── vite.config.ts                  # Vite build and plugin configuration

🚀 Installation & Local Setup

Prerequisites

Node.js: Version 20.x or higher
npm or bun / yarn / pnpm
Google Gemini API Key: Obtain a free key from Google AI Studio

1. Clone the Repository

git clone https://github.com/bikram73/Interview_Agent_AI.git
cd Interview_Agent_AI

2. Install Dependencies

npm install

3. Configure Environment Variables

Create a .env file in the root directory:

cp .env.example .env

Populate your Gemini API key:

GEMINI_API_KEY=your_actual_gemini_api_key_here

4. Run Development Server

npm run dev

The application will be live at http://localhost:3000.

5. Production Build

npm run build
npm run start

6. Netlify Deployment

This project is pre-configured with netlify.toml, public/_redirects, and /netlify/functions.

🔗 Live Production Site: https://interview-agent-ai.netlify.app/

Push code to your GitHub repository (https://github.com/bikram73/Interview_Agent_AI).
In the Netlify Dashboard, click Add new site > Import an existing project.
Build Settings will auto-detect from netlify.toml:
- Build command: npm run build
- Publish directory: dist
- Functions directory: netlify/functions
Add Environment Variable in Netlify (Site settings > Environment variables):
- GEMINI_API_KEY: your_gemini_api_key
Click Deploy Site!

🏗️ System Architecture & Workflow

+-----------------------------------------------------------------------------+
|                              CANDIDATE CLIENT                               |
|   Select Role Track -> Configure Experience -> Active Timed Session Room    |
|   (Speech-to-Text Dictation -> Text Area Input -> Submit Response)          |
+-----------------------------------------------------------------------------+
                                       |
                                       | POST /api/* (JSON)
                                       v
+-----------------------------------------------------------------------------+
|                       SERVER PROXY / NETLIFY FUNCTIONS                      |
|   - Express.js Routes (Local / Cloud Run) OR Netlify Serverless Functions   |
|   - Environment Variable Security (GEMINI_API_KEY hidden from browser)      |
|   - Input Sanitization & Payload Verification                               |
+-----------------------------------------------------------------------------+
                                       |
                                       | Google GenAI SDK
                                       v
+-----------------------------------------------------------------------------+
|                              GOOGLE GEMINI AI                               |
|   - Model: gemini-3.6-flash / gemini-2.5-flash                              |
|   - Strict JSON Schema Validation (Structured Outputs)                      |
|   - Temperature: 0.2 (Deterministic Rubric Assessment)                      |
+-----------------------------------------------------------------------------+
                                       |
                                       | Clean JSON
                                       v
+-----------------------------------------------------------------------------+
|                            EVALUATION & DOSSIER                             |
|   - Question Score (0-10) + Key Strengths + Gaps + Improvement Tip          |
|   - Aggregate Final Score (0-100%) + Panel Hiring Verdict                   |
|   - Candidate Downloadable Transcript (.txt)                                |
+-----------------------------------------------------------------------------+

For an in-depth breakdown of state transitions, data boundaries, and latency budgets, see ARCHITECTURE.md.

🧠 AI Prompting Strategy & Anti-Hallucination

The platform uses a four-pillar prompting framework to ensure accurate, constructive, and reproducible interview assessments:

Role-Bound System Persona: Gemini is prompted strictly as a Senior Principal Engineering & Hiring Committee Lead with explicit domain grading rubrics.
Schema-Constrained JSON Outputs: Output formatting uses Google GenAI SDK responseSchema (JSON Schema Type.OBJECT), preventing syntax errors, unescaped quotes, or conversational conversational filler.
STAR Rubric Grounding: Answers are evaluated against Situation, Task, Action, and Quantifiable Results. Generic claims without architectural context or metrics are penalized constructively.
Deterministic Fallback Engine: If any network anomaly occurs, the client includes deterministic, curated fallback questions and evaluations so the candidate's session is never interrupted.

📊 Confidence & Evaluation Scoring Formula

Each question is evaluated using an objective 10-point rubric:

$$\text{Question Score} = \mathbf{C}_{\text{tech}} (40%) + \mathbf{D}_{\text{depth}} (25%) + \mathbf{S}_{\text{comm}} (20%) + \mathbf{Q}_{\text{metrics}} (15%)$$

$\mathbf{C}_{\text{tech}}$ (Correctness & Accuracy): Technical accuracy, appropriate libraries, algorithms, and architectural patterns.
$\mathbf{D}_{\text{depth}}$ (Depth & Trade-offs): Awareness of edge cases, memory footprint, concurrency, security, and scalability.
$\mathbf{S}_{\text{comm}}$ (Clarity & Structure): Logical thought progression, clear technical vocabulary, and concise explanations.
$\mathbf{Q}_{\text{metrics}}$ (Impact & Quantitative Evidence): Tangible metrics, benchmark figures, and STAR-oriented results.

Aggregate Score & Verdict Tiers

90% – 100%: Strong Hire (Exceptional architectural depth and articulation)
75% – 89%: Recommended for Hire (Solid technical grounding with minor gaps)
60% – 74%: Borderline / Practice Needed (Foundational knowledge present, lacks quantifiable impact)
< 60%: Needs Fundamental Development (Significant conceptual errors or incomplete answers)

📄 Challenge Deliverables Checklist

Autonomous AI Question Generator: Generates custom technical and scenario questions for 11+ roles and custom inputs.
Live Candidate Dictation: Real-time voice speech-to-text integration with manual editing capability.
Instant AI Evaluation Engine: Provides numerical grading, strengths, weaknesses, and improvement roadmaps per question.
Final Hiring Panel Report: Complete executive summary, overall score, confidence ranking, and hiring verdict.
Exportable Dossier: Formatted plaintext download of full transcripts and evaluation notes.
Comprehensive API Documentation: Fully documented endpoints with request/response schemas.
Cloud & Serverless Deployment: Built for Express.js (Docker/Cloud Run) and Netlify Functions.

🖨️ Sample Structured JSON Output

1. Generated Question Sample

{
  "questions": [
    {
      "id": 1,
      "question": "How do Python decorators work under the hood, and how would you implement a decorator that measures function execution time?",
      "category": "Core Python Fundamentals",
      "difficulty": "Medium",
      "expectedKeyConcepts": [
        "First-class functions",
        "Closures and *args, **kwargs",
        "functools.wraps metadata preservation",
        "time.perf_counter"
      ]
    }
  ]
}

2. Answer Evaluation Sample

{
  "score": 8.5,
  "feedback": "Strong explanation demonstrating solid understanding of Python first-class functions, closures, and wrapper syntax.",
  "strength": "Accurately noted the usage of functools.wraps to preserve original function docstrings and metadata.",
  "weakness": "Did not mention how time.perf_counter() differs from time.time() for CPU benchmark precision.",
  "improvement": "Mention using time.perf_counter() for high-precision benchmarking and discuss handling asynchronous coroutines with async decorators.",
  "idealAnswerKeyPoints": [
    "Explain decorators as functions taking another function and returning a wrapper",
    "Preserve metadata using @functools.wraps(func)",
    "Capture start and end time using time.perf_counter()",
    "Return the original function result"
  ]
}

3. Final Report Sample

{
  "overallScore": 86,
  "strengths": [
    "Demonstrated mastery of language internals and asynchronous patterns",
    "Structured problem-solving approach adhering to STAR methodology",
    "Clear awareness of production error boundaries and unit testing"
  ],
  "weaknesses": [
    "Could incorporate more concrete benchmark figures when discussing performance optimization",
    "Expand further on distributed caching invalidation strategies"
  ],
  "recommendation": "Recommended for Hire",
  "topicsToImprove": [
    "Distributed Systems & Redis Cache Invalidation",
    "Asyncio Event Loop Internals",
    "STAR Method Quantifiable Impact"
  ],
  "confidence": "High",
  "summary": "The candidate performed exceptionally well across all 5 technical interview questions, demonstrating senior-level conceptual clarity and communicative composure."
}

🌐 REST API Documentation

1. `POST /api/generate-questions`

Generates a set of tailored interview questions.

Request Body:

{
  "role": "Python Developer",
  "experience": "Mid Level (3-5 yrs)",
  "question_count": 5
}

Response (200 OK):

{
  "questions": [
    {
      "id": 1,
      "question": "...",
      "category": "...",
      "difficulty": "...",
      "expectedKeyConcepts": ["..."]
    }
  ]
}

2. `POST /api/evaluate-answer`

Evaluates an individual answer against the specific interview question.

Request Body:

{
  "role": "Backend Developer",
  "question": "How do you prevent SQL injection in an Express and PostgreSQL application?",
  "answer": "I always use parameterized queries with pg or an ORM like Prisma/Drizzle instead of raw string concatenation.",
  "experience": "Junior (1-2 yrs)"
}

Response (200 OK):

{
  "score": 9.0,
  "feedback": "Excellent, direct answer correctly emphasizing parameterized queries.",
  "strength": "Identified parameterized queries as the gold standard for SQL injection prevention.",
  "weakness": "Could briefly mention input validation (e.g., Zod) as defense-in-depth.",
  "improvement": "Mention schema validation libraries like Zod alongside database-level escaping.",
  "idealAnswerKeyPoints": [
    "Parameterized queries / prepared statements",
    "ORMs with built-in parameterization",
    "Input validation and sanitization at the boundary"
  ]
}

3. `POST /api/generate-report`

Generates the comprehensive executive hiring report across the entire session.

Request Body:

{
  "role": "AI Engineer",
  "experience": "Mid Level (3-5 yrs)",
  "session": [
    {
      "questionIndex": 1,
      "question": "Explain RAG architecture and how you evaluate vector embeddings.",
      "candidateAnswer": "RAG retrieves context from vector DBs...",
      "score": 8.5,
      "feedback": "Good response.",
      "strength": "Clear explanation of chunking.",
      "weakness": "Could mention cosine similarity metrics.",
      "improvement": "Reference RAGAS evaluation framework."
    }
  ]
}

📄 Sample Documents & Extraction Results

To explore full sample transcripts, simulated user sessions, and evaluation breakdowns: 👉 Open SAMPLE_RESULTS.md for complete examples across:

🐍 Python Full-Stack Developer Session
🤖 AI & Machine Learning Engineer Session
🛡️ Cybersecurity Incident Response Session
👥 HR & Behavioral STAR Session

⚡ Performance Benchmarks & Latency

Operation	Model / Engine	Average Latency	Peak Memory
Question Generation (5 Qs)	`gemini-3.6-flash`	1.12s	~24 MB
Answer Evaluation	`gemini-3.6-flash`	0.84s	~18 MB
Final Dossier Generation	`gemini-3.6-flash`	1.35s	~28 MB
Client Bundle (Gzipped)	Vite + Rollup	—	~86 KB
Voice STT Processing	Web Speech API	Real-time (0ms)	Client Native

⚠️ Known Limitations & Recovery

Browser Speech Recognition: Web Speech API is natively supported in Chromium browsers (Chrome, Edge, Brave). Safari and Firefox candidates can type their responses directly with full keyboard accessibility.
API Rate Limiting: If Gemini API quotas are exceeded, the client seamlessly invokes local deterministic fallback questions and evaluations to ensure zero candidate disruption.
Microphone Permissions in iframes: When embedded in sandboxed iframes without mic permissions, candidates can use standard text input.

📄 License

This project is licensed under the MIT License — see the full LICENSE file for details.

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files to deal in the Software without restriction, including the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies.

_{Built with ❤️ using React 19, Tailwind CSS, TypeScript, and Google Gemini AI.}

Navigation Menu

Repository files navigation

🎙️ Interview Agent AI — Intelligent Mock Interview & Candidate Evaluation Platform

📑 Table of Contents

✨ Project Features & Capabilities

🛠️ Tech Stack & Languages

💻 Programming Languages

📦 Key Libraries & Frameworks

📂 File Structure

🚀 Installation & Local Setup

Prerequisites

1. Clone the Repository

2. Install Dependencies

3. Configure Environment Variables

4. Run Development Server

5. Production Build

6. Netlify Deployment

🏗️ System Architecture & Workflow

🧠 AI Prompting Strategy & Anti-Hallucination

📊 Confidence & Evaluation Scoring Formula

Aggregate Score & Verdict Tiers

📄 Challenge Deliverables Checklist

🖨️ Sample Structured JSON Output

1. Generated Question Sample

2. Answer Evaluation Sample

3. Final Report Sample

🌐 REST API Documentation

1. `POST /api/generate-questions`

2. `POST /api/evaluate-answer`

3. `POST /api/generate-report`

📄 Sample Documents & Extraction Results

⚡ Performance Benchmarks & Latency

⚠️ Known Limitations & Recovery

📄 License

About

Releases

Packages

Contributors

Languages

Name		Name	Last commit message	Last commit date
netlify/functions		netlify/functions
public		public
src		src
.env.example		.env.example
.gitignore		.gitignore
ARCHITECTURE.md		ARCHITECTURE.md
LICENSE		LICENSE
README.md		README.md
SAMPLE_RESULTS.md		SAMPLE_RESULTS.md
TECHNICAL_REPORT.md		TECHNICAL_REPORT.md
bun.lock		bun.lock
index.html		index.html
metadata.json		metadata.json
netlify.toml		netlify.toml
package.json		package.json
server.ts		server.ts
tsconfig.json		tsconfig.json
vite.config.ts		vite.config.ts

Latest commit

History

Folders and files

Repository files navigation

🎙️ Interview Agent AI — Intelligent Mock Interview & Candidate Evaluation Platform

📑 Table of Contents

✨ Project Features & Capabilities

🛠️ Tech Stack & Languages

💻 Programming Languages

📦 Key Libraries & Frameworks

📂 File Structure

🚀 Installation & Local Setup

Prerequisites

1. Clone the Repository

2. Install Dependencies

3. Configure Environment Variables

4. Run Development Server

5. Production Build

6. Netlify Deployment

🏗️ System Architecture & Workflow

🧠 AI Prompting Strategy & Anti-Hallucination

📊 Confidence & Evaluation Scoring Formula

Aggregate Score & Verdict Tiers

📄 Challenge Deliverables Checklist

🖨️ Sample Structured JSON Output

1. Generated Question Sample

2. Answer Evaluation Sample

3. Final Report Sample

🌐 REST API Documentation

1. POST /api/generate-questions

2. POST /api/evaluate-answer

3. POST /api/generate-report

📄 Sample Documents & Extraction Results

⚡ Performance Benchmarks & Latency

⚠️ Known Limitations & Recovery

📄 License

About

Topics

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages

1. `POST /api/generate-questions`

2. `POST /api/evaluate-answer`

3. `POST /api/generate-report`