Build¶

Hands-on projects and the patterns production AI systems are made of. 16 pages. Generated by scripts/build_manifest.py - do not edit by hand.

Projects¶

Page	In one line	Reviewed
Build a Chatbot in an Afternoon	Build a working, streaming chatbot with a web UI in about 60 lines of Python, ground it in your own document, and measure what each conversation costs.	2026-10-21
Build a Claude Agent with MCP	Build a small Claude agent with one custom MCP server that answers questions over files and a SQLite database, in about 30 minutes.	2026-10-21
Build a RAG Pipeline	Build an end-to-end RAG pipeline (chunk, embed, store in pgvector, retrieve, generate with Claude, evaluate) in about 30 minutes.	2026-10-21
Build a Structured Extraction Pipeline	Turn messy text (invoices, emails, contracts) into validated, typed records with a schema, a cheap model, a gold test set and the batch API, and know your field-level accuracy before anyone relies on the data.	2026-10-21
Build a Voice Agent	Build a push-to-talk voice assistant from three swappable parts (speech-to-text, an LLM, text-to-speech), measure where the latency goes, and learn when to switch to a speech-to-speech model instead.	2026-10-21
Deploy a Machine Learning Model	Train a simple classical ML model, register it, and deploy it to a managed inference endpoint with monitoring, in 3-4 hours.	2026-10-21
Fine-Tune with LoRA	LoRA-fine-tune a small open-weights model on a tiny dataset, then prove with an eval whether it beat the base model.	2026-10-21
Red-Team Your Own App for Prompt Injection	Build a small tool-using assistant, attack it with a scored suite of prompt injections hidden in documents, then add layered defenses and prove with numbers which ones actually helped.	2026-10-21
Run Llama on a Single GPU	Stand up a self-hosted, OpenAI-compatible endpoint for an open-weights Llama model on one GPU with vLLM.	2026-10-21
Set Up an Eval Harness	Build a small but real eval harness (golden set, automated runs, regression detection, CI gate) so quality drops show up on the pull request, not in user complaints.	2026-10-21

Patterns¶

Page	In one line	Reviewed
Advanced RAG Patterns	When basic "embed, retrieve top 5, answer" stops being good enough, these patterns (better chunking, hybrid search, reranking, query rewriting, contextual retrieval, GraphRAG and agentic retrieval) fix specific, diagnosable failures; measure first, then add only the one that fixes your failure.	2026-10-21
Agent Architectures	An agent is an LLM calling tools in a loop until a task is done; this page shows the six common shapes (from a single tool loop to multi-agent orchestration), when each one earns its extra cost, and where to put the controls that keep it safe.	2026-10-21
AI/ML Pipeline Architecture	Design an end-to-end ML pipeline (data, features, training, registry, serving, monitoring) on AWS, Azure or GCP using MLOps practices.	2026-10-21
Cost and Latency Control for LLM Apps	Measure cost per completed task and time to first token, then pull the levers in order (free wins like caching, batching and trimming tokens first; trade-offs like smaller models, lower effort and routing second) and re-check quality after every change.	2026-10-21
LLM Application Architecture	A production LLM app is a normal web service with five extra layers (a gateway, prompt and context assembly, the model call, output checks, and an evaluation and feedback loop), and most failures happen in those layers, not in the model.	2026-10-21
LLMOps and Observability	Run LLM features like any production service (traces, metrics, alerts, versioned releases) plus three LLM-specific loops: offline evals before every change, online quality monitoring after, and a path from user feedback back into the eval set.	2026-10-21