At a Glance

Name: Build an AI Icebreaker Bot with IBM Granite 3.0 & LlamaIndex
Price: Free CAD
Rating: 5.0 (6 reviews)
Author: ricky_shi, boyun_leung, joseph_santarcangelo, hailey_quach

Use RAG with LlamaIndex, IBM watsonx models (including Granite 3.0), and ProxyCurl API to gather insights from LinkedIn profiles, enabling the generation of a personalized icebreaker bot with career insights, and more. The system extracts fun facts, career highlights, and personal interests to craft unique conversation starters. Ideal for networking, job interviews, or social events, this icebreaker bot helps you make a strong impression with customized insights, making your introductions more engaging and memorable in any professional or social setting.

Imagine this: You’re heading to a high-profile networking event, one filled with industry leaders, potential employers, and future collaborators. You’ve done your research—sort of. You’ve scrolled through LinkedIn, trying to learn more about the people you’ll be meeting, but the profiles are overwhelming, and your nerves are kicking in. You need to make a good first impression, but how can you stand out when everyone is likely preparing the same generic questions and introductions?

Now, picture this instead: You walk into the event feeling confident, armed with an AI-powered tool that has already done the hard work for you. You type a name into your phone, and in seconds, the bot scours LinkedIn gathering relevant facts—personal interests, recent career milestones, skills and projects. The bot, powered by LlamaIndex and IBM watsonx models (including Granite 3.0), generates a set of personalized icebreakers, designed just for this moment.

Instead of starting with the usual "So, what do you do?" you lead with something unique: "I saw you recently spoke at a conference about AI ethics. That must have been an incredible experience—what’s one key takeaway you’ve had from those conversations?" Immediately, you’ve caught their attention. It’s personal, informed, and shows you’ve done your homework—except, you didn’t have to, because your AI assistant did it for you.

The AI icebreaker bot turns the stress of introductions into an opportunity to shine. By leveraging advanced data extraction techniques and natural language processing (NLP), it gathers meaningful insights and generates tailored conversation starters, helping you break the ice with anyone—whether it’s a potential employer, a colleague, or even at a social gathering.

This is the problem the AI icebreaker bot solves. It’s not just about finding facts—it’s about helping you connect with people on a deeper level, whether you’re navigating a professional event or making new friends. By the end of this project, you’ll have built an AI-driven tool that removes the guesswork from introductions, empowering you to make memorable first impressions wherever you go.

Source: DALL-E

What you’ll learn

By completing this lab, you will gain valuable skills to:

Access LinkedIn profile data with the ProxyCurl API
Learn how to use the ProxyCurl API to efficiently retrieve LinkedIn profile data, overcoming the challenges of scraping profiles directly due to web scraping restrictions. This skill will allow you to extract key information such as work experience, education, and more from LinkedIn profiles.
Split and process JSON data for indexing
Discover how to split JSON data from LinkedIn profiles into smaller chunks, or nodes, for easy processing. You’ll learn to use tools like LlamaIndex for efficiently dividing profile data into manageable parts, enabling more granular insights, and retrieval.
Use watsonx embeddings to represent data as vectors
Learn how to use IBM watsonx.ai models for embedding text data into vector representations. This is a crucial step for transforming LinkedIn profile information into vector embeddings, which allows for more accurate and efficient information retrieval in AI applications.
Create a vector database using LlamaIndex
Store and index document chunks (nodes) in a VectorStoreIndex using LlamaIndex, ensuring that the LinkedIn profile data can be retrieved efficiently based on similarity to user queries. You’ll learn how vector indexing enhances the speed and accuracy of data retrieval for AI-driven applications.
Query and retrieve data using Prompt Engineering
Master the art of prompt engineering by designing prompts that extract key career facts from the indexed data and answer user-specific queries. You’ll use LlamaIndex’s custom PromptTemplate to define detailed prompts that ensure high-quality responses from the IBM Granite 3.0 model.
Build a simple, interactive chatbot interface
Implement a conversational chatbot interface that allows users to ask in-depth questions about the LinkedIn profiles, providing detailed responses generated by AI. You’ll simulate real-world conversational AI interactions for personalized icebreakers, career insights, and more.

Why are these skills essential for AI-driven data processing?

In today’s digital and professional landscape, the ability to quickly retrieve and analyze data from social media profiles, like LinkedIn, is increasingly important. Whether for networking, job interviews, or professional interactions, having access to personalized icebreakers and insights can give you an edge. Manually collecting and analyzing this data can be slow and error-prone, but leveraging tools like LlamaIndex, IBM Granite, and ProxyCurl allows for automated data collection, processing, and retrieval.

This project introduces you to key AI concepts like Retrieval-Augmented Generation (RAG) and vector databases that are critical for modern AI applications, including personalized assistants, chatbots, and LLM-powered tools. By using advanced embedding and vectorization techniques, you’ll be able to convert raw text into meaningful insights in an efficient and scalable way. These techniques allow you to automate the process of finding relevant information, making it accessible for AI-powered applications.

Who should complete this lab?

This lab is perfect for:

AI and Machine Learning enthusiasts looking to integrate watsonx.ai and LlamaIndex into their applications to create smart, scalable tools.
Data Scientists who want to automate data retrieval and build efficient AI systems for processing social media or professional networking data.
Developers building AI-powered applications or chatbots that rely on extracting data and providing personalized insights.

What you’ll need

Before starting this lab, ensure you have the following:

Basic knowledge of Python programming
Familiarity with Python will help you navigate the data retrieval, data processing, and AI querying steps in this project.
Familiarity with APIs and data structures
Understanding APIs (like ProxyCurl) and data structures (like JSON) will make it easier to grasp the process of accessing and handling profile data.
A current version of a web browser
To run the project and test the chatbot interface, you’ll need a web browser like Chrome, Edge, Firefox, or Safari.

Get started with the AI-powered icebreaker bot

Start this guided project today and learn how to use LlamaIndex, IBM watsonx models, and ProxyCurl to build an AI-powered icebreaker bot that generates personalized conversation starters from LinkedIn profiles. By the end of the lab, you’ll have the skills to build AI-driven applications that can handle data extraction, data processing, and intelligent querying to deliver personalized insights.

Offered By: IBMSkillsNetwork

Build an AI Icebreaker Bot with IBM Granite 3.0 & LlamaIndex