AdaptIQ

An adaptive, AI-powered technical interview simulator. Upload a resume, and the app extracts your skills, indexes it with FAISS for semantic retrieval, conducts a context-aware mock interview that adapts question difficulty to your answers, and evaluates each response in real time.

Built with Streamlit, LangChain, Groq (Llama 3.3 70B), and FAISS.

---

Features

• Resume parsing — Upload a PDF resume; text is extracted and chunked automatically.
• Semantic indexing — Resume chunks are embedded using all-MiniLM-L6-v2 and indexed with FAISS; a similarity search retrieves the most relevant context (skills, projects, experience) before each interview question.
• Skill extraction — Technical skills are pulled directly from resume content using an LLM, with no hardcoded skill lists.
• Adaptive interview — Each question is generated from the FAISS-retrieved resume context and your full answer history, so the interviewer can increase difficulty after strong answers, probe deeper, or ease off after weak ones.
• Real-time evaluation — Every answer is critiqued (technical accuracy, communication, depth) immediately after submission.
• Full session history — All questions, answers, and evaluations are logged and viewable in an expandable history panel.
• Modern dashboard UI — Dark-themed SaaS-style layout with a sidebar for upload/progress and a main workspace for skills, interview flow, and evaluation results.

No skills, questions, or scores are hardcoded anywhere in the app — every value displayed comes from the backend LLM functions.

---

Project Structure

.
├── app.py                     # Streamlit frontend — orchestrates the full interview flow
├── utils/
│   ├── llm.py                 # Shared Groq LLM client (Llama 3.3 70B via LangChain)
│   ├── pdf_parser.py          # PDF text extraction + chunking
│   ├── skill_extractor.py     # LLM-based skill extraction from resume text
│   ├── interviewer.py         # Adaptive question generation based on resume context + history
│   ├── evaluator.py           # LLM-based answer evaluation
│   └── vector_store.py        # FAISS vector store — embedding, indexing, and context retrieval
├── .env                        # Holds GROQ_API_KEY (not committed)
└── README.md

---

How It Works

1. Upload & Process — The resume PDF is parsed (pdf_parser.extract_chunks) into full text and overlapping chunks.
2. Semantic Indexing — Chunks are embedded with sentence-transformers/all-MiniLM-L6-v2 and stored in a FAISS index (vector_store.create_vectorstore). A similarity search against "skills projects experience technologies" retrieves the top 5 most relevant chunks, joined into a compact resume_context (vector_store.get_resume_context).
3. Skill Extraction — The full resume text (not the excerpt) is sent to the skill extractor (skill_extractor.extract_skills) to ensure no skills are missed, returning a list rendered as pills in the UI.
4. Start Interview — The first question is generated (interviewer.generate_question) using resume_context as context. No prior history exists yet, so the question is grounded purely in the retrieved resume excerpt.
5. Submit Answer — Your answer is evaluated (evaluator.evaluate_answer) against the current question, returning free-form feedback on accuracy, communication, and depth.
6. Next Question — The question/answer/evaluation triple is appended to session history, and the next question is generated using resume_context and the entire conversation so far — making each subsequent question adaptive to how you've performed.
7. Repeat — Steps 5–6 continue for as long as you want. All state is held in Streamlit's session_state, so progress persists across reruns within a browser session.

---

Tech Stack

Layer	Technology
Frontend	Streamlit
LLM orchestration	LangChain
LLM inference	Groq — llama-3.3-70b-versatile
PDF parsing	pypdf
Text chunking	LangChain RecursiveCharacterTextSplitter
Vector store	FAISS via langchain_community
Embeddings	sentence-transformers/all-MiniLM-L6-v2 via HuggingFaceEmbeddings

---

Setup

1. Clone and install dependencies

git clone <your-repo-url>
cd <project-folder>
pip install -r requirements.txt

Required packages:

streamlit
python-dotenv
langchain
langchain-groq
langchain-community
langchain-text-splitters
pypdf
faiss-cpu
sentence-transformers

Note: sentence-transformers will download the all-MiniLM-L6-v2 model (~80 MB) on first run. Subsequent runs use the cached model.

2. Configure your API key

Create a .env file in the project root:

GROQ_API_KEY=your_groq_api_key_here

Get a free key at console.groq.com/keys.

3. Run the app

streamlit run app.py

The app will open at http://localhost:8501.

---

Usage

1. Open the app and upload a resume PDF from the sidebar.
2. Click Process Resume — the resume is parsed, semantically indexed with FAISS, and extracted skills appear as pills in the main panel.
3. Click Start Interview to receive your first question, grounded in the most relevant parts of your resume.
4. Type your answer and click Submit Answer to see your evaluation.
5. Click Next Question to receive an adaptive follow-up based on your resume context and performance so far.
6. Review the full question/answer/evaluation history at any point in the Interview History panel.
7. Use Reset Session in the sidebar to clear progress and start over with a new resume.