Complete AI Mastery Course

{"contentFormat":"slides.v1","completion":{"requireAllSlides":true,"requireQuiz":true},"slides":[{"kind":"title","eyebrow":"Module 8: Portfolio Projects","title":"Project 1: Building a Personal AI Knowledge Base","body":"Turn your documents into a private, searchable AI assistant that answers questions with citations. In 60–90 minutes you'll build a working knowledge base using free tools (as of 2026). No coding required.","outcomes":["Explain what a Personal AI Knowledge Base is and why it matters","Describe the three core components: documents, retrieval, chat interface","Understand RAG and embeddings in plain language","Build a knowledge base with NotebookLM or Claude Projects","Test and verify answers with citations","Document your project for your portfolio"],"narration":"Welcome to your first portfolio project. By the end of this lesson, you'll have a working AI knowledge base that you can talk to in plain language and that answers from your own documents with citations. Let's get started."},{"kind":"content","heading":"What Is a Personal AI Knowledge Base?","body":"A knowledge base is an organised collection of documents. A Personal AI Knowledge Base adds an AI layer that reads your documents and answers questions grounded in their actual content, not just its general training. Ask \"What did my insurance policy say about flooding?\" and it quotes the clause instead of guessing.\n\nWhy it's valuable:\n- Saves search time\n- Reduces errors\n- Keeps your knowledge private and reusable\n- Demonstrates a skill employers and clients care about in 2026: turning messy documents into an AI assistant that knows your context","callout":{"variant":"insight","title":"Real-World Impact","text":"A university student in Nairobi loads 20 lecture PDFs and past exam papers into one notebook. Before exams she asks, 'Summarise the key arguments on monetary policy from weeks 4–6, with the slide they came from,' and studies from grounded, cited notes."},"narration":"A Personal AI Knowledge Base is like hiring a personal researcher who has already read everything you own. It saves time, reduces errors, and keeps your information private."},{"kind":"content","heading":"The Three Core Components","body":"Every Personal AI Knowledge Base has three parts:\n\n1. Documents (your sources) — the raw material: PDFs, notes, spreadsheets, transcripts, web pages. Quality in, quality out.\n2. Retrieval (finding the right pieces) — the system that pulls the relevant parts of your documents to answer a question.\n3. A chat interface — where you ask questions in natural language and read grounded answers.","callout":{"variant":"tip","title":"Start Small","text":"Pick one focused theme — e.g., 'My AI course notes,' 'Freelance contracts,' or 'Immigration paperwork.' A focused base gives better answers."},"narration":"Three components work together: your documents, a retrieval system that finds relevant pieces, and a chat interface where you ask questions."},{"kind":"content","heading":"RAG and Embeddings, Explained Simply","body":"The technique that powers this is RAG (Retrieval-Augmented Generation). Here's the plain-English version:\n\n- Your documents are split into small chunks.\n- Each chunk is turned into an embedding — a list of numbers that captures its meaning. Think of it as giving every paragraph a GPS coordinate on a 'map of meaning,' so paragraphs about similar topics sit close together.\n- When you ask a question, your question also becomes a coordinate. The system grabs the nearest chunks (retrieval) and hands them to the AI, which writes an answer using them (generation).\n\nAnalogy: Instead of the AI answering from memory (like a student guessing on an exam), RAG is an open-book exam — the AI is handed the exact pages first, then answers. That's why it's more accurate and can cite sources.\n\nYou do not need to build RAG from scratch. Modern tools do it for you.","narration":"RAG is like an open-book exam. Your documents are split into chunks, each given a meaning coordinate. When you ask a question, the system finds the nearest chunks and hands them to the AI to write an answer. You don't need to build this yourself — modern tools handle it."},{"kind":"content","heading":"Real Tool Options (Free Tiers as of 2026)","body":"Choose one tool for your first build. All have free tiers:\n\n- NotebookLM (Google) — purpose-built for this. Upload sources (PDF, Word, Markdown, CSV, Google Docs, web URLs, YouTube links, EPUB, and more), then chat with automatic citations. Free: up to 50 sources per notebook.\n- Claude Projects (Anthropic) — upload documents to a project's knowledge base plus custom instructions. Free accounts get up to 5 projects. Paid plans add larger-scale RAG retrieval.\n- ChatGPT (OpenAI) — upload files into a chat or a custom GPT and ask questions across them.\n- Low-code / no-code stack — combine a note app (Notion, Obsidian) with an AI connector. Optional and more advanced.\n\nFor your first build, NotebookLM or Claude Projects is the fastest path.","callout":{"variant":"warning","title":"Privacy & Security","text":"Avoid uploading sensitive personal information (e.g., ID numbers, medical records) to cloud services. For sensitive data, use a local tool or check the provider's data handling policy."},"narration":"You have several free tool options. NotebookLM and Claude Projects are the fastest for your first build. Remember to avoid uploading sensitive personal information to cloud services."},{"kind":"content","heading":"Hands-On: Build It (Step by Step)","body":"Budget 60–90 minutes. Follow these steps:\n\n1. Choose a focused topic. Pick one useful theme — e.g., 'My AI course notes,' 'Freelance contracts,' or 'Immigration paperwork.'\n\n2. Gather 5–15 documents. Collect clean sources: PDFs, Word/Markdown files, or web links. Rename them clearly (2026-tax-guide.pdf, not scan001.pdf).\n\n3. Create the knowledge base.\n- NotebookLM: go to the site, create a new notebook, click Add source, and upload your files or paste URLs.\n- Claude Projects: create a new Project, then add files to Project knowledge. In custom instructions, write: \"Answer only from the uploaded documents. If the answer isn't there, say so. Always name the source.\"\n\n4. Test with three question types. Ask:\n- A factual question: \"What is the deadline mentioned for [X]?\"\n- A summary question: \"Summarise the main points of the onboarding document in 5 bullets.\"\n- A cross-document question: \"Do any two documents contradict each other on [topic]? Quote both.\"\n\n5. Verify every answer. Click the citation or ask \"Show me the exact sentence you used.\" Open the source and confirm.\n\n6. Add a 'house rules' prompt to reduce errors:\n> \"You are my research assistant. Answer strictly from my documents. Quote the source name and the relevant line. If the documents don't contain the answer, reply: 'Not found in your sources.' Never guess.\"\n\n7. Organise for growth. Keep a simple index note listing each document and what it covers. Update when you add sources.\n\n8. Document the project — this is what makes it portfolio-ready.","callout":{"variant":"warning","title":"Watch Your Usage Limits","text":"Free tiers have limits (e.g., NotebookLM: 50 sources per notebook; Claude Projects: 5 projects). If you exceed them, you may need to upgrade to a paid plan. Check current pricing on the provider's website."},"narration":"Follow these eight steps to build your knowledge base. Start with a focused topic, gather 5-15 documents, create the base, test with three question types, verify answers, add a house rules prompt, organise for growth, and document your project."},{"kind":"content","heading":"What Good Looks Like: Checklist","body":"Before you call this project done, check that you have:\n\n- [ ] A knowledge base with at least 5 documents on one focused topic\n- [ ] Tested factual, summary, and cross-document questions\n- [ ] Verified at least 3 answers by checking the cited source\n- [ ] Added a 'house rules' prompt that prevents guessing\n- [ ] Created an index note listing your documents\n- [ ] Written a short project summary (2–3 paragraphs) covering:\n - What you built and why\n - Which tool you used and why\n - One challenge you faced and how you solved it\n - One thing you would improve next time\n\nExample summary (for a freelance developer in Brazil):\n> \"I built a personal AI knowledge base for my freelance contracts and project briefs using NotebookLM. I chose it because it's free and provides automatic citations. A challenge was that some PDFs had scanned images — I had to convert them to text first. Next time, I'll add client emails to make the base even more useful.\"","narration":"Use this checklist to ensure your project is complete. The most important part is the written summary — that's what makes it a portfolio piece."},{"kind":"quiz","heading":"Check Your Understanding","questions":[{"question":"What does RAG stand for?","options":["Retrieval-Augmented Generation","Random Access Generation","Real-time AI Guidance","Retrieval-Assisted Grounding"],"questionId":"cmrf73kc0002vpd27pyg31eby"},{"question":"Which of the following is a privacy best practice when building a personal AI knowledge base?","options":["Upload all documents including those with personal ID numbers","Avoid uploading sensitive personal information to cloud services","Share your knowledge base publicly for feedback","Use the same password for all your accounts"],"questionId":"cmrf73kc0002wpd27s6hkyblr"},{"question":"What should you do if the AI cannot find an answer in your documents?","options":["Let the AI guess based on its training","Reply 'Not found in your sources' as instructed in the house rules prompt","Upload more documents until it finds something","Ignore the error and move on"],"questionId":"cmrf73kc0002xpd27pl7pghci"}],"quizId":"qz_cmk7ln1qv003fg4p86ffljvza"},{"kind":"summary","heading":"Project 1: Key Takeaways","takeaways":["A Personal AI Knowledge Base turns your documents into a searchable, chat-based assistant that answers with citations.","RAG (Retrieval-Augmented Generation) powers this by retrieving relevant chunks and generating grounded answers.","Free tools like NotebookLM and Claude Projects make building one easy and fast (60–90 minutes).","Always verify answers by checking the cited source.","Add a 'house rules' prompt to prevent guessing and ensure accuracy.","Document your project with a short summary to showcase in your portfolio."],"narration":"You now have a working Personal AI Knowledge Base. Remember to verify answers, use house rules, and document your project. This is a strong portfolio piece that demonstrates practical AI skills."}]}