Quick reference: AI Engineering foundations
Broader context. This roadmap focuses on AI security, but the topics below are tested in AI Engineer interviews and strengthen your ability to reason about what you're attacking and defending. Read these as supplementary — they're not gated by any week.
| Topic | Why it matters for security | Where to start |
|---|---|---|
| LoRA / QLoRA / PEFT | Adversarial fine-tuning, adapter backdoors, supply chain poisoning | Hugging Face PEFT docs; Sebastian Raschka's LoRA explainer |
| Structured Outputs / JSON mode | Output parsing exploits, schema injection, type confusion | OpenAI Structured Outputs guide; Instructor library |
| LLM Evaluation & Benchmarks | Red team eval frameworks, measuring guardrail effectiveness | HELM benchmark; lm-evaluation-harness; DeepEval |
| Inference Optimization | Latency-based side channels, resource exhaustion DoS, serving vulnerabilities | vLLM docs; TensorRT-LLM; GGUF/llama.cpp quantization |
| Function Calling / Tool Use | Tool poisoning, schema injection, SSRF via tools | Anthropic tool use docs (covered in Week 5) |
| Multi-modal models | Vision jailbreaks, image-based injection, audio adversarial | LLaVA paper; GPT-4V red team report (covered in Week 10) |
Recommended reading for AI engineering breadth:
- 📖 Designing Machine Learning Systems (Chip Huyen, O'Reilly) — ML systems design, feature stores, monitoring, production patterns
- 📖 Build a Large Language Model (From Scratch) (Sebastian Raschka, Manning) — already in Week 3 resources; tokenization through fine-tuning
- 📖 AI Engineering (Chip Huyen, O'Reilly 2025) — LLM application architecture, RAG, agents, evaluation, deployment
- 📄 Latent Space podcast — weekly interviews with AI engineers; good for staying current on production patterns
- 📄 The AI Engineer newsletter (Swyx) — curated industry signal on tooling, benchmarks, deployment