Loading…
MixedMediumAI EngineeringArtificial Intelligence
AI Engineering with LLMs: Beginner
Build and test a small LLM feature in Python: a bilingual tutorial with one running project, EvidenceDesk, a graded lab, hand-picked videos, and three exams. Runs offline on a fake model; no API key, GPU or paid service.
20 lessons3 exams15h 25m
About this course
Level 1, path 1 (part 1 of 2) of the BitByte AI path. Calling a model once is easy; building a feature that cites evidence, refuses when it should, survives bad output and stays testable is engineering. This course teaches that on one small project, EvidenceDesk, starting from a provider-neutral client and ending with a packaged, documented and tested application.
You get a written tutorial in English and Bangla, hand-picked videos, and three exams: two quizzes and a harder final exam of code reading, output prediction, debugging and design choices. The tutorial also has a graded practice lab and a capstone with a grader and an acceptance script. Everything runs offline against a fake model and toy embeddings, so no API key, GPU or paid service is needed. The course is educational: facts were checked on 12 October 2026, model names, prices and SDK details change, and the Bangla text was written by AI. It makes no promise about jobs or results.
What you will learn
- Design a provider-neutral LLM client with adapters, configuration, timeouts and controlled errors, and test it with a fake client
- Assemble context with clear trust boundaries and a token budget, and make calls reproducible
- Handle streaming and validate structured outputs, with controlled errors for malformed or incomplete results
- Build embeddings, similarity and local vector search, and judge retrieval quality on a labelled fixture set
- Build a small RAG system with chunking, validated citations and abstention when evidence is missing
- Register and call tools safely: argument validation, permissions and failure handling
- Write evaluation datasets, deterministic tests and regression gates, and count cost and latency with budgets and caching
- Defend against prompt injection with access boundaries and safe-failure tests, then package and document the project reproducibly
Curriculum
1. Read: AI Engineering with LLMs — Beginner
1 item2. LLM APIs and a Provider-Neutral Client
2 items3. From Prompt to Context Engineering
2 items- What Is Context Engineering? Why It Matters for AI Agents9:56
- What is a Context Window? Unlocking LLM Secrets11:30
4. Check yourself: API and context basics
1 item5. Streaming, Structured Outputs, and Reliable Responses
2 items- Structured Output from LLMs: Grammars, Regex, and State Machines17:20
- Python Tutorial: Type Hinting vs Type Checking vs Data Validation - What's the Difference?18:16
6. Embeddings and Local Vector Search
2 items- Text embeddings & semantic search3:30
- What is a Vector Database? Powering Semantic Search & AI Applications9:48
7. First RAG System: Chunking, Retrieval, and Citations
2 items- What is Retrieval-Augmented Generation (RAG)?6:35
- RAG Evaluation: Precision, Recall, Faithfulness, RAGAS Explained Clearly12:19
8. Tool and Function Calling
1 item- What is Tool Calling? Connecting LLMs to Your Data4:56
9. Check yourself: building LLM features
1 item10. Basic Evals and Regression Testing
2 items- Error Analysis: The Highest ROI Technique In AI Engineering12:08
- LLM as a Judge: Scaling AI Evaluation Strategies6:08
11. Cost, Latency, and Caching Fundamentals
2 items- Most devs don't understand how LLM tokens work10:57
- What is Prompt Caching? Optimize LLM Latency with AI Transformers9:06
12. Basic AI Application Security
2 items- What Is a Prompt Injection Attack?10:57
- OWASP's Top 10 Ways to Attack LLMs: AI Vulnerabilities Exposed25:11
13. Check yourself: final exam
1 item14. Full-length courses (optional)
3 items- [1hr Talk] Intro to Large Language Models59:48
- Please Learn How To Write Tests in Python... Pytest Tutorial33:09
- Packaging Your Python Code With pyproject.toml | Complete Code Conversation51:39
Requirements
- Python 3.12, a terminal, and basic Python: functions, classes, JSON, virtual environments and running tests (Level 0 covers this)
- No API key, GPU or paid account: everything runs offline with a fake client and toy embeddings
- Comfortable reading English or Bangla (the tutorial and exams are in both)
Reviews
Sign in to join the discussion.