TalentMatch

TalentMatch is an AI-powered resume shortlisting and candidate matching system built with Retrieval Augmented Generation (RAG). It helps recruiters instantly find the best-fit candidates from hundreds of resumes by understanding both semantic meaning and exact keywords.

Live Demo: https://talentmatch-ai-assistant.streamlit.app/

Portfolio: https://prachi-gore-portfolio.netlify.app/

Features

• Multi-Format Support - Parses PDF, DOCX, TXT, and Markdown resumes
• Intelligent Parsing - Extracts skills, experience, and metadata using LLM
• Hybrid Vector Search - Dense embeddings (semantic) + TF-IDF (keywords)
• Production Vector DB - Stores hybrid vectors in Pinecone for scalability
• Smart Reranking - Cross-encoder improves ranking precision after retrieval
• Structured Output - Pydantic-validated JSON for reliable API/UI integration
• Stateful Conversations - Chat memory for follow-up questions without re-retrieval

Tech Stack

• Python
• Streamlit
• LangChain
• OpenAI
• Pinecone
• Sentence Transformers
• Numpy

Project Structure

talentmatch/
  config.py            # Configuration & environment settings
  ingestion.py         # Resume ingestion, parsing, embedding pipeline
  retrieval.py         # Hybrid search, query expansion, reranking
  prompts.py           # YAML prompt templates & Pydantic parser
  schemas.py           # Pydantic data models
  chains.py            # LangChain orchestration
  chat.py              # Session management & memory
  llm.py               # LLM interface (OpenAI GPT-4)
  pinecone_client.py   # Vector database client
scripts/
  ingest_resumes.py    # CLI: Index resumes to Pinecone
  chat_cli.py          # CLI: Interactive chat
app.py                 # Streamlit web dashboard
prompt.yaml            # LLM prompt template

Setup

python -m venv venv
venv\Scripts\activate
pip install -r requirements.txt
copy .env.example .env

Add your API keys inside .env:

OPENAI_API_KEY=your_openai_key
PINECONE_API_KEY=your_pinecone_key

For Pinecone serverless, the default values are:

PINECONE_CLOUD=aws
PINECONE_REGION=us-east-1

Run TalentMatch Dashboard

streamlit run app.py

Open the local URL shown in the terminal, usually:

http://localhost:8501

How It Works

Resume files
  -> text extraction
  -> LLM resume parsing
  -> dense embeddings + sparse TF-IDF
  -> Pinecone hybrid index
  -> query expansion
  -> hybrid retrieval
  -> cross-encoder reranking
  -> structured LLM response
  -> chat memory
  -> Streamlit UI

The first user message performs resume retrieval and shortlisting. Follow-up questions are answered from the stored shortlist for that session.

Example Query

Shortlist the top 5 Python backend developers with FastAPI, AWS, and SQL experience.

Follow-up:

Which candidate has the strongest cloud experience?