Learn AI · Browse tracks
Core Concepts
AI and LLM basics — what AI can do, classical ML literacy, tokens, attention, embeddings, prompting, RAG, agents, evals, multimodal, open vs closed models, privacy, and a mini-GPT you can train.
How this relates Browse here for Hello Interview–style articles. The guided path below in the sidebar remains the project milestone sequence.
Articles in this track
- What is AI (and what it is not)Machine learning vs deep learning vs generative AI — vocabulary, limits, and when not to use an LLM.Open
- Classical ML literacyTrain/val/test splits, leakage, overfitting, and why baselines still matter in an LLM world.Open
- TokenizationTokens are the interface between UTF-8 text and model IDs — BPE, costs, chat templates, and why bad splits break products.Open
- Self-attentionAttention lets every token look at every other token — Q/K/V, multi-head, causal masks, and why context length is expensive.Open
- Embeddings and similarityVectors that put “nearby meaning” nearby in space — foundation of semantic search, RAG, and clustering.Open
- Prompt engineering fundamentalsPrinciples from DeepLearning.AI’s prompt course — clear instructions, delimiters, iterative refinement, and when to stop prompting and build systems.Open
- Structured outputsJSON schemas, tool calls, and validation loops — making LLMs produce machine-checkable results.Open
- RAG building blocksChunking, hybrid retrieval, rerank, citations, and evals — the grounded-generation system before frameworks.Open
- Agents and the ReAct loopAn agent is an LLM in a loop over tools and state — what it is, how products use it, architecture, how to build one, today’s stacks, and failure modes.Open
- Evals fundamentalsOffline golden sets, online monitors, component vs E2E, judges, and ship gates before demos.Open
- Serving and streamingFrom generate() to production APIs — TTFT, tokens/sec, SSE streaming, batching, and OpenAI-compatible façades.Open
- Networking for AI appsTimeouts, retries, idempotency, websockets vs SSE, and how LLM latency changes API design.Open
- LLM project lifecycleScope → model choice → adapt (prompt/RAG/fine-tune) → evaluate → deploy → monitor — the GenAI-with-LLMs lifecycle for product teams.Open
- Multimodal basicsImages, audio, and documents into LLMs — when modality helps, how APIs expose it, and failure modes (OCR, resolution, cost).Open
- Privacy and data for AI appsPII, retention, tenant isolation, and what not to put in prompts, logs, or vector indexes — the AI-for-Everyone society week applied to builders.Open
- Open-weight models vs closed APIsWhen to call OpenAI/Anthropic vs run Llama/Mistral yourself — cost, latency, data control, and capability tradeoffs.Open
- Mini-GPT from scratchNext-token prediction at laptop scale — tokens, causal attention, train until loss drops, reload your checkpoint. The same loop frontier models use.Open