AI Fluency & Building · AI Knowledge Systems
Go deeper. Real-world tasks that need more research, planning and a polished result.
Task 1 of 3
Answer customer questions with sources.
What to do
Build a RAG chatbot over a business's FAQs, policies and product pages (a brand you choose) that answers customer questions and links to the page it used for every answer.
You’ll make
Done looks like
A chatbot that cites the source page for every answer.
Task 2 of 3
Find which one finds the right answer most often.
What to do
Test 3 ways of splitting documents (fixed size, by heading, by paragraph) on the same business content. Measure which retrieves the right passage most often across 20 customer questions.
You’ll make
Done looks like
Results across 20 questions showing which strategy works best.
Task 3 of 3
Design the fields and load 100 records.
What to do
Design a metadata schema for a brand's content library (type, topic, product, audience, date, language) and load 100 records into a vector database.
You’ll make
Done looks like
A metadata schema and 100 records loaded into a vector database.
Learn before you build
Free course
A free course on transformers, embeddings and building with LLMs.
Free tool
A free open-source vector database that runs on your laptop.
Article guide
Clear explainers on embeddings, chunking and vector search.
Article guide
A practical technique that sharply reduces failed retrievals.
Article guide
A framework for connecting LLMs to your own data.
Article guide
Build a question-answering app over your own documents.