GEO for Developers: Prompt Engineering


Prompt engineering ensures AI consumes structured, context-rich data. Combine:


Knowledge graphs


Product metadata


Embeddings



Exercise: Build a mini RAG pipeline:


1. Convert pages to structured JSON



2. Generate embeddings



3. Query an LLM



4. Evaluate responses for relevance and accuracy



GEO for Developers: Fine-Tuning


Fine-tuning LLMs aligns AI outputs with brand voice and accuracy. Instruction-tuned models produce consistent and relevant responses.


Example:


Fine-tune an LLM on product FAQs for Lafz lipsticks.


AI now consistently answers queries like: “Which Lafz lipstick is best for dry skin?”



Exercise:


Collect FAQs from 5 products.


Fine-tune a small LLM and test for consistent outputs.




---



GEO for Developers: RAG Pipelines


Retrieval-Augmented Generation (RAG) combines AI generation with structured retrieval.


Steps:


1. Store content in a vector database.


2. Query relevant vectors based on user input.



3. Feed results into the LLM for generation.


Demo:


AI recommends products based on user queries like “Top matte lipsticks under $20.”


Vector search ensures accurate, semantically relevant recommendations.


Exercise:


Build a mini RAG system for 10 products and evaluate AI responses.


-Case Study: E-Commerce


Generative engines enhance e-commerce:


Product recommendations


Semantic search


Summarized reviews


Example:


An online store in Dhaka integrates AI to recommend outfits based on past purchases.


Customers see full sets instead of individual items.


Exercise:


Take 5 products and generate AI-recommended bundles.


Compare AI results vs manual recommendations.


---


Case Study: News Portals


AI can summarize articles, extract entities, and suggest related content.


Example:


Bangladeshi news portal uses AI to summarize daily news and highlight top trending entities.


Readers get concise summaries and related content suggestions.


Exercise:


Pick 5 recent articles. Generate AI summaries and entity extractions.


Evaluate accuracy and readability.

Case Study: Enterprise Documentation


Large enterprises use GEO for internal knowledge management:


AI Q&A over manuals


Knowledge graphs for fast search


Semantic retrieval for onboarding



Demo:


Build a small internal documentation database.


Query AI for common technical questions and verify accuracy.



Exercise:


Create a knowledge graph for 10 internal processes and test AI comprehension.