A Resume–Job Fit Analysis chatbot built using Gradio, FAISS Vector Search, and the Hugging Face Inference API.
This Space hosts the GenAI Career Agent, a generative AI that analyzes user resumes, retrieves structured resume data through a vectorstore (FAISS), and evaluates how well a candidate fits any provided job description.
🚀 Features
- AI Career Coach – Helps users understand job fit, strengths, and areas for improvement.
- RAG Pipeline – Uses FAISS to retrieve relevant resume chunks.
- LLM-Powered Analysis – Uses a remote Hugging Face model via
InferenceClient.
- Structured JSON Output including:
job_fit_scorefit_summarystrengthsmissing_skillsrecommendations
- Secure Token Handling with Hugging Face Space Secrets.
📌 Current Capability
✔ Resume Parsing & Analysis
The system currently parses the user's resume (pre-embedded with MiniLM) and produces job-fit analytics using RAG + LLM inference.
🛠️ Upcoming Features
🔜 GitHub Repo Intelligence
- Automatic retrieval of repositories
- Summarization of project impact
- Extraction of tech stack & coding patterns
- Integration into the job-fit score
🔜 LinkedIn Profile Integration
- Work history extraction
- Skill inference
- Keyword alignment
- Soft-skill assessment
These features will be integrated into the same RAG pipeline so the model can reason across Resume + GitHub + LinkedIn for a unified career profile.
🧠 How It Works
- Resume data is pre-embedded using
sentence-transformers/all-MiniLM-L6-v2.
- Embeddings are stored inside
data/vectorstores/.
- The FAISS retriever fetches the most relevant resume sections based on the job description.
- A custom prompt formats the retrieved text and sends it to the LLM.
- The LLM generates structured JSON insights.
🗂️ Tech Stack
- Gradio 5 (ChatInterface front-end)
- LangChain Runnables
- FAISS Vector Search
- HuggingFace Embeddings
- Hugging Face Inference API
🔒 Token Handling
Set secrets in your Space
📜 Model Licensing & Notices
🧍 Personal / Educational Use
This is a personal project, intended solely for educational and career-analysis purposes.
Users are solely responsible for how they use the outputs.
