Agentic Engineering Bootcamp

First principles for engineers who govern what agents build.

Five structured programmes that take a competent software engineer to a competent agentic engineer - someone who can develop, verify, evaluate, and govern agent-native systems.

I: Linux Substrate - the kernel primitives, shell, filesystem, containers, and networking that agents actually run on. 12 steps, 51-65 hours.

II: Agentic Engineering Practices - how to work with agents: architecture, context engineering, verification, the human-AI interface, governance. 11 steps, 50-61 hours.

III: Operational Analytics - the 20% of data science that does 80% of the heavy lifting. pandas, DuckDB, statistics, visualisation, cost modelling. 10 steps, 32-40 hours.

IV: Evaluation & Adversarial Testing - how do you know if it’s working? Eval design, LLM-as-judge, agent evaluation, red teaming, safety. 9 steps, 39-48 hours.

V: Agent Infrastructure in Practice - RAG, state management, conversation memory, observability, debugging, production patterns. 9 steps, 36-45 hours.

Total: 51 steps, ~208-259 hours. Not a weekend project - a practitioner’s field guide built from operational experience.

Ranking criteria for the learning order within each bootcamp:

  1. Compositional leverage - does this knowledge compose into everything above it?
  2. Return per hour - how much capability per unit of learning time?
  3. Irreplaceability - can an agent compensate for your ignorance, or must you understand it?

Step I.1 is available now as a sample. The rest is in development.

I: Linux Substrate

51-65 hours

1. The Unix Process Model 4-6 hours 2. The Shell Language 8-12 hours 3. The Filesystem as State 4-6 hours 4. The Text Processing Pipeline - grep, sed, awk, jq, yq 6-8 hours 5. Python as a CLI Tool - The Escape Hatch from Shell Complexity 4 hours 6. Make and Just as Orchestrators 4-5 hours 7. Git Internals - Beyond Porcelain Commands 4 hours 8. Process Observation - strace, ltrace, lsof, ss 3 hours 9. Container Internals - Namespaces, Cgroups, and OverlayFS 4 hours 10. Networking from the CLI - curl, openssl, dig, tcpdump, iptables 4 hours 11. Process Supervision - systemd, cron, and supervisord 3 hours 12. Advanced Bash - trap, coprocesses, associative arrays, BASH_REMATCH, and production patterns 3 hours

II: Agentic Engineering Practices

50-61 hours

1. How LLMs Actually Work 5-6 hours 2. Agent Architecture Patterns 4-5 hours 3. Prompt Engineering as System Design 5-6 hours 4. Context Engineering 5-6 hours 5. Tool Design and Agent-Computer Interfaces 4-5 hours 6. Verification and Quality for Probabilistic Systems 6-7 hours 7. The Human-AI Interface 5-6 hours 8. Multi-Model Verification Strategies 3-4 hours 9. Failure Modes and Recovery 4-5 hours 10. Governance, Process, and Enterprise Integration 5-6 hours 11. Cost, Security, Legal, and Compliance 4-5 hours

III: Operational Analytics

32-40 hours

1. Tabular Data with pandas and DuckDB 4-5 hours 2. Descriptive Statistics 3-4 hours 3. SQL Analytics with DuckDB 4-5 hours 4. Statistical Testing for Practitioners 3-4 hours 5. Time Series Basics 3-4 hours 6. Visualisation for Decision-Making 4-5 hours 7. Log Analysis Patterns 3-4 hours 8. Cost Modelling for API-Based Systems 3-4 hours 9. Text Analysis Basics 3-4 hours 10. Notebook-Based Analysis Workflows 2-3 hours

IV: Evaluation & Adversarial Testing

39-48 hours

1. What Evaluations Actually Measure 4-5 hours 2. Designing Eval Datasets and Success Criteria 5-6 hours 3. Scoring and Grading Methods 5-6 hours 4. Evaluating Agents and Workflows 5-6 hours 5. Eval Infrastructure and Automation 4-5 hours 6. Adversarial Testing Methodology 5-6 hours 7. Red Teaming for Safety-Critical Capabilities 5-6 hours 8. Interpreting and Communicating Eval Results 3-4 hours 9. Building an Eval Culture 3-4 hours

V: Agent Infrastructure in Practice

36-45 hours

1. The Retrieval Problem 3-4 hours 2. Embeddings and Vector Search 4-5 hours 3. RAG Pipeline Engineering 5-6 hours 4. Advanced Retrieval Patterns 4-5 hours 5. State Management for Agents 4-5 hours 6. Conversation Memory and Session Persistence 4-5 hours 7. Agent Observability and Tracing 4-5 hours 8. Debugging Agent Systems 4-5 hours 9. Production Patterns 4-5 hours