How to Build Your Own GenAI Apps with FOSS Tools

Generative AI is transforming industries, but most people assume you need access to expensive closed APIs like GPT-4 or Gemini to start building. This session breaks that myth by showing how Free and Open Source Software (FOSS) empowers developers to build their own GenAI applications in minutes.

We’ll cover the basics of a GenAI workflow—from prompting, to Retrieval-Augmented Generation (RAG), to multi-step agents—and show how the open-source ecosystem offers powerful, free alternatives at every stage.

Popular OSS stacks we’ll highlight:

• Models: LLaMA 2, Mistral, Gemma (open weights for text generation)

• Vector Databases: FAISS, Chroma (for retrieval & semantic search)

• Frameworks: LangChain, Haystack (for chaining models, building RAG pipelines)

To make the session practical, I’ll do a live demo: “Chat with Your PDF” using Hugging Face Transformers + FAISS. In just a few lines of code, attendees will see how to query their own documents with an open-source model—no closed APIs, no cost barriers.

This talk is aimed at developers who want to get hands-on with open-source GenAI, understand the building blocks, and leave inspired to experiment further.

Key Takeaways from this talk:

• Understand the core GenAI workflow: Prompting → RAG → Agents.

• Discover the open-source stack for building GenAI apps (models, vector DBs, frameworks).

• Learn how to set up a basic GenAI application locally with Hugging Face + FAISS.

• Realize that you don’t need proprietary APIs—in 30 minutes, anyone can build a working AI-powered app using FOSS tools.

• Get inspired to contribute back to the ecosystem through code, datasets, or tutorials.