AI Engineer

2 h kickoff + 48 sessions × 2 h + 2 h showcase = 100 hours

1. Kickoff2 hoursSunday, November 1

   Kickoff and setup

   Meet the cohort, see how classes run, set up your development environment and accounts, tour what you will build, and make a first working call to an AI model.

2. Week112 hoursNov 2–7

   Building with language models

   1. How language models work, for builders: tokens, context and sampling
   2. Prompt design and system instructions
   3. Structured outputs and schemas
   4. Calling model APIs from code
   5. Cost, speed and choosing a model
   6. Project setup, version control and environments for AI apps

   Project A structured-extraction tool that turns messy text into clean data.

3. Week212 hoursNov 9–14

   Tools and workflows

   1. Tool and function calling
   2. Multi-step workflows and prompt chaining
   3. Streaming, retries and error handling
   4. Wrapping a model in a backend API
   5. Testing code that depends on a model
   6. Connecting to third-party services

   Project A tool-using assistant that runs behind its own API.

4. Week312 hoursNov 16–21

   Retrieval, part 1

   1. Embeddings and vector search
   2. Chunking documents and building an index
   3. Combining keyword and vector search
   4. Answers with citations
   5. A document question-and-answer app
   6. Working with real, messy documents

   Project Question answering over a real set of documents.

5. Week412 hoursNov 23–28

   Retrieval, part 2

   1. Reranking and query rewriting
   2. Metadata filters and access control
   3. Tables, PDFs and other hard inputs
   4. Measuring retrieval quality
   5. Keeping an index fresh
   6. Knowing when retrieval is the wrong tool

   Project Improve last week's app and show the measured gain.

6. Week512 hoursNov 30–Dec 5

   Agents

   1. The agent loop: plan, act, observe
   2. Choosing and describing tools
   3. State and memory
   4. Multi-agent patterns, and when to avoid them
   5. Limits, approvals and human-in-the-loop
   6. Recovering from failure

   Project A task agent with safeguards on what it can do.

7. Week612 hoursDec 7–12

   Evaluation

   1. Building test sets from real examples
   2. Automated checks and model-graded evaluations, and their limits
   3. Human review workflows
   4. Regression tests for prompts and pipelines
   5. Tracing and logging
   6. Comparing models and prompts fairly

   Project An evaluation harness for your retrieval app or agent.

8. Week712 hoursDec 14–19

   Reliability, safety and cost

   1. Latency, caching and rate limits
   2. Fallbacks and graceful failure
   3. Prompt injection and other attacks
   4. Privacy and sensitive data
   5. Monitoring and alerts in production
   6. Controlling cost

   Project Harden your project for real users.

9. Week812 hoursDec 21–26

   Shipping and capstone

   1. Adding an AI feature to an existing codebase
   2. Code review, documentation and handoff
   3. Deployment
   4. Capstone build sprint, part 1
   5. Capstone build sprint, part 2
   6. Demo preparation and course review

   Project Capstone: an AI feature built end to end.

10. Showcase2 hoursSunday, December 27

    Final showcase

    Present your capstone to the group, get feedback from trainers and classmates, and plan what to build and learn next.