Prompt Engineering
Turn a desired outcome into clear, testable instructions and output contracts. 8 course placements; 8 published now.
← Back to Compute College
Compute College program
Build AI systems that work outside the demo.
Learn how to design, evaluate, secure, operate, and optimize AI systems — from the first instruction to the production workload.
A five-course path with 48 published lesson placements. Learn prompts, context, evaluation, agents, and production workload economics in one practical sequence.
The program
Five courses, one practical sequence
AI Engineering starts with instructions and context, then moves through evaluation, controlled autonomy, and the economics of operating a real workload.
Context & Retrieval Engineering
Select, organize, protect, and manage the information available to a model. 10 course placements; 10 published now.
AI Evaluation & Reliability
Determine whether an AI application is useful, stable, safe, and ready to operate. 8 course placements; 8 published now.
Agent & Tool Engineering
Design controlled model workflows that use tools and external actions safely. 10 course placements; 10 published now.
AI Workload Engineering
Connect application requirements to model choice, infrastructure, reliability, and cost per outcome. 12 course placements; 12 published now.
Choose a route
Start where your work is
Builder path
Prompt Engineering → Context & Retrieval Engineering → AI Evaluation & Reliability → Agent & Tool Engineering → AI Workload Engineering.
Product and operations path
Prompt Engineering → AI Evaluation & Reliability → AI Workload Engineering, with selected context and agent lessons as your work requires them.
Course outline
Learn the system in layers
Every lesson opens directly. Follow the sequence at your own pace; no account or model API is required to follow the program.
Prompt Engineering
1. What is prompt engineering?
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
2. Define the AI task and success criteria
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
3. Prompt anatomy: instructions, context, and output
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
4. Clear instructions and constraints
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
5. Few-shot examples and counterexamples
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
6. Structured outputs and output contracts
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
7. Decomposition, workflows, and prompt chaining
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
8. Prompt versioning and regression testing
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
Context & Retrieval Engineering
1. What is context engineering?
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
2. Context windows and token budgets
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
3. What belongs in context?
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
4. Retrieval-augmented generation explained
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
5. Chunking and document preparation
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
6. Ranking, reranking, and retrieval quality
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
7. Long context is not the same as good context
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
8. Conversation history, state, and memory
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
9. Context compression, caching, and reuse
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
10. Untrusted context and prompt injection
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
AI Evaluation & Reliability
1. Why AI applications need evaluations
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
2. Define the task, ground truth, and scoring rubric
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
3. Build a practical evaluation dataset
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
4. Deterministic graders and rule-based checks
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
5. Human review and model-based graders
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
6. Evaluate retrieval and grounded answers
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
7. Evaluate agents and multi-step workflows
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
8. Production reliability and regression gates
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
Agent & Tool Engineering
1. What is an AI agent?
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
2. When not to build an agent
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
3. The agent stack: model, instructions, tools, harness, and environment
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
4. Tool design and function calling
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
5. Agent loops, stopping conditions, and budgets
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
6. Workflows before agents
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
7. Single-agent vs multi-agent design
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
8. Human approval and reversible actions
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
9. Agent security and excessive agency
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
10. Agent observability and evaluation
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
AI Workload Engineering
1. What is an AI workload?
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
2. Define functional and nonfunctional requirements
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
3. Workload patterns
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
4. Model selection for the workload
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
5. Hosted API vs managed deployment vs self-hosting
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
6. Latency: TTFT, output speed, and end-to-end response time
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
7. Throughput, concurrency, batching, and queues
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
8. Token, context, and cache economics
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
9. Routing, fallbacks, and model portfolios
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
10. Reliability and operational design
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
11. Observability and GenAI operations
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
12. Cost per successful outcome
Read the live introductory lesson and connect the concept to model quality, tokens, latency, and compute cost.
Practice the concepts
Use the AI Engineering tools
These browser-local exercises turn the lessons into numbers and decisions. They use illustrative assumptions and do not send your entries to a server.
AI Workload Design Scorecard
Turn a proposed AI system into a workload and architecture checklist.
Context Budget Calculator
Reserve output, then see how instructions and retrieved context use the window.
Prompt Evaluation Worksheet
Compare accepted results, retries, and review time across a fixed test set.
RAG Retrieval Exercise
Explore retrieval precision, recall, and context-token volume.
AI Workload Planner
Plan token economics from traffic and first-attempt success.
Agent Permission Designer
Classify an action before granting an agent access to it.
Cost per Successful Task Calculator
Include failures and review in the cost of an accepted result.
First implementation slice
Begin with the published foundations
Start with any of the 48 lessons, then use the course sequence and browser-local tools to connect design choices to production workload economics.
What is prompt engineering?
Turn a desired outcome into clear, testable instructions and output contracts.
What is context engineering?
Understand the instructions, history, retrieved information, tools, and state available at inference time.