Engineering guide
AI-Native Engineering
How specifications, verification, context, review, and technical ownership change when AI participates in software delivery.
Open guideLearn
Understand the technical issue before choosing the service or architecture.
These curated guides help business owners, technical decision makers, developers, researchers, and students understand software ownership, AI-system architecture, mobile computer vision, and model engineering. Start with the question closest to your project, then follow the relevant service and evidence.
Free guided courses
AI is part of our future. Learning it should be open to everyone.
Start with the basics, explore HerbDev examples, and check your understanding as you go. Free lessons, audio and practice scores. No account needed.
AI 101
Understand prompts, evidence, agents and the practical systems around a model.
6 lessons · 18 questions · Start AI 101 →
ML 101
Explore data, training, evaluation and the evidence behind a useful model.
6 lessons · 18 questions · Start ML 101 →
Choose what to learn next — practical projects, focused review and deeper guides.
Need custom courses for your company?
Teach employees your operating systems, approved AI tools and everyday workflows, with guided lessons built around your team.
Discuss custom employee courses →Vocabulary
Need a quick definition? Explore AI, machine learning, mathematics, and programming terms, with examples and connections between concepts.
Browse the vocabulary →A useful learning path should lead to a decision.
- 1
Understand
Clarify the technical issue and the constraints around it.
- 2
Connect
Find the service responsible for solving that class of problem.
- 3
Evaluate
Review relevant engineering evidence and its limitations.
- 4
Act
Discuss the project or request an assessment when the next step is clear.
Learning path 1
Application Rescue & Software Ownership
How do you assess, recover, and take ownership of software that is incomplete, fragile, or difficult to release?
Understand technical debt, AI-generated code, handoffs, production readiness, and the difference between repairing a system and rebuilding it.
Decision guide
AI Built My App—Now What?
Understand why generated code can look complete while still lacking validation, release readiness, and accountable ownership.
Open guideDecision guide
Fix the Existing App or Rebuild?
Frame the evidence needed to compare repair, selective replacement, and a controlled rebuild.
Open guideOwnership guide
Taking Over an Existing Codebase
Learn what a responsible handoff must establish across code, accounts, builds, deployment, documentation, and responsibility.
Open guideLearning path 2
AI Systems & Architecture
How do models, retrieval, tools, agents, guardrails, and human approvals become an operable system?
Move beyond a model demo and understand the architecture, control boundaries, evaluation, and workflow design around practical AI.
Start here
How AI Works
A plain-English path from a user request through tokens, model processing, safety checks, and a response.
Open guideAgent systems
The Agent Loop
How an agent plans, uses tools, observes results, and decides what to do next.
Open guideProduction architecture
AI Agent Best Practices
Scope, permissions, evaluation, monitoring, failure handling, and human review for tool-using systems.
Open guideRetrieval architecture
GraphRAG for Complex Data
How graph structure and grouped evidence can support questions that span connected information.
Open guideLearning path 3
Mobile, Computer Vision & On-Device AI
Should intelligence run on a device, in the cloud, or across both—and what does the image evidence really support?
Understand offline operation, privacy, latency, model and hardware limits, cross-platform vision tools, and responsible visual inference.
Architecture decision
Embedded AI vs Cloud AI
Compare local, remote, and hybrid inference around workflow, cost, privacy, connectivity, and maintenance.
Open guidePlatform guide
iOS Vision vs Android Computer Vision
Compare native vision frameworks, embedded models, classical computer vision, and normalized cross-platform output.
Open guideArchitecture comparison
Waymo vs Tesla Architectures
See how sensing strategy, world representation, validation, and operating constraints shape a vision system.
Open guideLearning path 4
Machine Learning & Model Engineering
How are models trained, evaluated, selected, optimized, and validated for a real operating environment?
Use model size, data quality, memory, evaluation, deployment constraints, and failure cases to make a defensible engineering choice.
Model selection
Large Models, Small Models, and the System Around Them
Understand how runtime location, serving economics, training, memory, and hybrid routing shape model choice.
Open guideEngineering experiments
Neural-Network Training Experiments
Ten reproducible examples covering baselines, controlled optimization, validation evidence, and limitations.
Open guideTraining strategy
How Big Models Teach Small Models
Knowledge distillation, teacher/student models, soft labels, and the path toward smaller deployable models.
Open guideRuntime engineering
Why LLM Memory Gets Expensive
How context, KV cache, batch size, and decoding affect memory and deployment economics.
Open guideFrom understanding to action
Where does the uncertainty live in your system?
Share the codebase state, workflow, data, target devices, deployment constraints, or model decision. HerbDev can help identify whether the next step is an assessment, rescue, architecture decision, or focused experiment.
Discuss Your ProjectManage browser learning data
“Remove my test data” clears all AI 101 and ML 101 scores, reviews, project checklists and rating selections, plus saved vocabulary and recent history in this browser. Test entries cannot be separated from other learning progress. Cookie preferences, security protections, submitted feedback, contact emails and past analytics are kept.