Overview
You build one thing: an assistant for the support desk at Quillpath, a project-management software company. It triages tickets, drafts replies from the help centre for a person to review, and handles refund requests by a written policy, never moving money without a person approving. Every technique in the course is measured on the same 36 tickets, so you see exactly what it changes. Quillpath is fictional and its data is simulated; every lesson also opens with a real, cited case.
What you will learn
- 103 written parts, yours for good
- 34h 48m of reading, measured not estimated
- Written for the intermediate level
- Every future revision included
- UPI, cards and netbanking
Prerequisites
- A Google account, to run code in Google Colab. Nothing to install.
- One free API key, from Groq or Google Gemini, to use in Colab. Or no key at all, if you run a model on your own computer with Ollama and work in a code editor.
- For a few lessons (a container, a GPU serving engine, a CI pipeline), your own computer. Those lessons say so.
Curriculum
16 sections · 103 parts · 34h 48m3h 34m free to read · 31h 14m with the course
01Course guide5 parts · 62 min
- Large Language Models: The Complete CoursePreview5 min
- How This Course WorksPreview5 min
- Setup: Colab, Your Model Provider, and Running LocallyPreview15 min
- The Project: the Quillpath Support DeskPreview14 min
- Shared Quillpath DataPreview23 min
02Module 1: What a Large Language Model Is8 parts · 152 min
- Module 1: What a Large Language Model IsPreview1 min
- Lesson 1: A model that only predicts the next tokenPreview24 min
- Lesson 2: What pretraining gave the model, and what it costsPreview23 min
- Lesson 3: How a text predictor becomes an assistantPreview16 min
- Lesson 4: Measuring what a model can and cannot doPreview24 min
- Lesson 5: Choosing a model when the landscape keeps movingPreview21 min
- Lab: Start the support desk packagePreview31 min
- Incident: A reading list of books that do not existPreview12 min
03Module 2: Tokens, Context, and Cost7 parts · 141 min
- Module 2: Tokens, Context, and Cost1 min
- Lesson 1: Tokens: what the model reads and bills22 min
- Lesson 2: What fits in the window, and what the model uses21 min
- Lesson 3: Keeping a long thread inside its budget27 min
- Lesson 4: What a ticket costs, and how to pay less25 min
- Lab: Count, budget and price every call30 min
- Incident: A timestamp that switched the cache off15 min
04Module 3: Inference and Decoding6 parts · 109 min
- Module 3: Inference and Decoding1 min
- Lesson 1: How a model writes its reply, one token at a time23 min
- Lesson 2: How the model picks each word, and the dials19 min
- Lesson 3: Forcing a format, and paying for thinking28 min
- Lab: Put decoding settings in the package31 min
- Incident: A sampling bug turns replies into word salad7 min
05Module 4: Prompt Engineering7 parts · 144 min
- Module 4: Prompt Engineering1 min
- Lesson 1: What a prompt is made of, and what the model reads22 min
- Lesson 2: Build the test set before you tune a prompt23 min
- Lesson 3: Examples, wording and personas, measured on tickets21 min
- Lesson 4: Change one thing, log it, and know when to stop25 min
- Lab: Prompt files, a frozen split and a harness39 min
- Incident: Same models, three harnesses, three leaderboards13 min
06Module 5: Reasoning and Advanced Prompting7 parts · 131 min
- Module 5: Reasoning and Advanced Prompting1 min
- Lesson 1: Ask the model to show its working21 min
- Lesson 2: Ask several times, then vote or check19 min
- Lesson 3: Split the work into steps you can check23 min
- Lesson 4: Fix answers with a check, tune prompts with tests26 min
- Lab: Decide refunds by reading, voting and rule32 min
- Incident: AI explainers publish wrong interest sums9 min
07Module 6: Structured Outputs and Tool Use6 parts · 117 min
- Module 6: Structured Outputs and Tool Use1 min
- Lesson 1: Answers your code can trust22 min
- Lesson 2: Let the model ask your code to act20 min
- Lesson 3: Control tool calls and distrust what the model asks24 min
- Lab: Typed triage, five tools and their guards39 min
- Incident: A chatbot invents its owners' qualifications11 min
08Module 7: Context Engineering7 parts · 153 min
- Module 7: Context Engineering1 min
- Lesson 1: Why more context can make answers worse24 min
- Lesson 2: Put the right help articles in front of the model26 min
- Lesson 3: Packing every request to a token budget23 min
- Lesson 4: Memory the support desk can explain and erase21 min
- Lab: Retrieval, a budgeted context and memory48 min
- Incident: Hidden text steered a coding assistant10 min
09Module 8: Agents7 parts · 148 min
- Module 8: Agents1 min
- Lesson 1: Let the model choose the next step, and measure it25 min
- Lesson 2: Guards that stop, retry, pause and record an agent22 min
- Lesson 3: Give the agent new reach, and fence what it can touch20 min
- Lesson 4: Several agents, and grading the path they took27 min
- Lab: One traced refund agent, with every guard46 min
- Incident: An agent paid out far more than asked7 min
10Module 9: Fine-Tuning and Adaptation7 parts · 172 min
- Module 9: Fine-Tuning and Adaptation1 min
- Lesson 1: Should you fine-tune? Decide, then measure the bar24 min
- Lesson 2: Build training data you can trust29 min
- Lesson 3: Train SmolLM2 two ways and score the test once27 min
- Lesson 4: Teaching preferences, then checking what broke25 min
- Lab: The fine-tuning trial, from plan to decision49 min
- Incident: Reworded test questions passed the leak check17 min
11Module 10: Evaluation8 parts · 187 min
- Module 10: Evaluation1 min
- Lesson 1: Build the eval before the feature28 min
- Lesson 2: Noise, sample size and fair comparisons22 min
- Lesson 3: Human evaluation: two raters, one rubric, one number18 min
- Lesson 4: A model as judge, checked against people26 min
- Lesson 5: Evaluate the whole system, then keep it honest26 min
- Lab: An eval suite that gates every change57 min
- Incident: Launch charts that disagreed with their numbers9 min
12Module 11: Safety, Security, and Alignment7 parts · 129 min
- Module 11: Safety, Security, and Alignment1 min
- Lesson 1: What alignment fixes, and what the model remembers25 min
- Lesson 2: Map where the assistant can be attacked18 min
- Lesson 3: Defences against injection, each one measured23 min
- Lesson 4: Deploying responsibly: privacy, fairness, policy, disclosure18 min
- Lab: Defences in code, and a red-team set37 min
- Incident: A chatbot reply exposed support agents' sessions7 min
13Module 12: Multimodal Models6 parts · 108 min
- Module 12: Multimodal Models1 min
- Lesson 1: Reading screenshots with vision-language models19 min
- Lesson 2: Reading invoice PDFs and scans you can trust18 min
- Lesson 3: Voice messages, screen recordings and choosing a design20 min
- Lab: One checked intake for every attachment41 min
- Incident: Fake receipts that read perfectly9 min
14Module 13: Deployment, Operations, and Economics7 parts · 138 min
- Module 13: Deployment, Operations, and Economics1 min
- Lesson 1: Where the model runs, and one door in front18 min
- Lesson 2: Running the model yourself, and what it costs22 min
- Lesson 3: Keeping the assistant fast, up and affordable24 min
- Lesson 4: Seeing what the model did, and changing it safely22 min
- Lab: One gateway, a web service and a way back43 min
- Incident: A leaked key ran up 82,314 dollars8 min
15Module 14: Applications, Patterns, and Frontiers6 parts · 109 min
- Module 14: Applications, Patterns, and Frontiers1 min
- Lesson 1: Seven proven patterns, one support desk24 min
- Lesson 2: Design for the day the model is wrong18 min
- Lesson 3: Knowing when not to use the model20 min
- Lab: Review placement, feedback and a limitations memo38 min
- Incident: Generated stories published with blank slots8 min
16Module 15: Capstone2 parts · 88 min
- Module 15: Capstone1 min
- Lab: Capstone: one track, evidence a reviewer trusts87 min
Part 1 · freeLarge Language Models: The Complete Course · 5 min
What a large language model is, how it reads and writes, how to make it reliable, how to test it, how to secure it and how to run it, measured on one real-shaped task from the first lesson to the last.
You build one thing: an assistant for the support desk at Quillpath, a project-management software company. It triages tickets, drafts replies from the help centre for a person to review, and handles refund requests by a written policy, never moving money without a person approving. Every technique in the course is measured on the same 36 tickets, so you see exactly what it changes. Quillpath is fictional and its data is simulated; every lesson also opens with a real, cited case.
Who this course is for
You can write Python: functions, classes, dictionaries, running a notebook. You do not need a machine learning background.
Reviews
No reviews yet. The first comes from a reader who finishes the course.
to review this course once you have finished it.
Next to this one
Transformers: From Attention to Trained, Served and Interpreted Models111 parts · 43h 12m · ₹399Retrieval-Augmented Generation: The Complete Course95 parts · 38h 33m · ₹399Forward Deployed Engineer: The Complete Course85 parts · 18h 15m · ₹399AI Engineering: The Complete Course105 parts · 23h 44m · ₹399Agentic AI Engineer: The Complete Course97 parts · 33h 10m · ₹399
Large Language Models80 parts · 27h 0m · Free
Model Context Protocol83 parts · 14h 44m · ₹199
RAG82 parts · 13h 36m · ₹399
Python79 parts · 7h 28m · Free