🎓 Career Companion is an interactive GenAI-powered assistant designed to help college students with personalized career guidance, course suggestions, and tech pathway insights — all powered by cutting-edge generative AI tools and hosted on Kaggle Notebooks.

🚀 Project Overview

As part of my Generative AI Capstone Project, I built a Career Companion Assistant that simulates interactive Q&A sessions for college students, helping them explore the right career choices and resources in real-time using Large Language Models (LLMs).

The assistant supports both chat-based interaction and form-driven recommendations, making it beginner-friendly and effective even without real-time UI deployment.

🧠 Key GenAI Capabilities Demonstrated

This project integrates three core GenAI capabilities:

✅ 1. Few-shot Prompting

  • The assistant uses prompt engineering to provide consistent responses with career suggestions, even for ambiguous or brief questions.
  • Includes career role templates (e.g., software engineer, data analyst, cloud engineer) with predefined suggestions and courses.

✅ 2. Structured Output (Controlled Generation)

  • Responses are formatted in a consistent, structured way (lists, markdown tables, and sections).
  • Useful for logging into CSV files and analyzing student needs.

✅ 3. Retrieval Augmented Generation (RAG)

  • Simulated document lookup and embeddings are used to enrich the assistant's answers.
  • Students can get domain-specific insights like tech stacks, roadmap links, and certifications.

🧰 Tools & Technologies Used

  • 🧠 Gemini Pro (via API) – for LLM-based Q&A
  • 🐍 Python & Pandas – for backend logic and chat logging
  • 📊 Kaggle Notebooks – as the main interactive environment
  • 📄 CSV Export – to save Q&A logs for further analysis
  • 🔗 Markdown display – to show responses in clean UI format

💡 How It Works (with Screenshots)

The project runs inside a Kaggle Notebook with the following components:

  1. Model Initialization

    • Sets up Gemini Pro API and context for responses.

  2. Prompt Handling Function

    • A function handles user prompts (dummy input block on Kaggle) and logs responses.

  3. Response Generator

    • Calls the model and uses prompt templates for consistent responses.

  4. Chat Logging

    • Stores every Q&A in a Pandas DataFrame and saves it as career_chat_log.csv.

  5. Download Link

    • Allows users to download the full session as a log file from Kaggle.

📷 Screenshots

  • Setup block with Gemini Pro key

Setup block with Gemini Pro key

  • Sample Q&A block in notebook

Sample Q&A block in notebook

  • Downloadable log file

Downloadable log file

  • Structured response with markdown

Structured response with markdown

🔍 Sample Output

Sample Output

🔗 Try the Project

You can view the full notebook here:

👉 Kaggle Notebook Link

📚 What I Learned

  • How to use LLMs with few-shot prompting
  • Creating structured logs for better analysis
  • Simulating interaction in a non-UI environment like Kaggle
  • Basics of Retrieval Augmented Generation (RAG)

🛠 Improvements Planned

  • Integrate embeddings using FAISS/ChromaDB
  • Use Streamlit or Gradio for web UI
  • Allow real-time user input with context memory

✅ GenAI Capabilities Recap

Capability Used
✅ Structured Output ✅
✅ Few-shot Prompting ✅
✅ RAG ✅

🙌 Final Thoughts

GenAI isn't just about fancy chatbots — it can truly empower students by providing context-aware, structured, and actionable guidance. This project gave me real-world insight into building domain-specific assistants using the power of generative AI.

If you're a student looking to get into AI, education tech, or LLM apps — start experimenting today!

📬 Feel free to connect or ask questions in the comments!

LinkedIn
GitHub
Twitter
Kaggle