Learn to design, build, and ship real AI systems — from ML/deep learning fundamentals through transformers and LLMs, prompt engineering, agents and multi-agent systems, RAG, fine-tuning, deployment, and AI safety.
21 modules and 5 complete real-project builds, with worked calculations, a downloadable kit, and a certificate of completion.
How AI, machine learning, and deep learning nest, narrow vs. general AI, and the AI engineering lifecycle.
Supervised vs. unsupervised learning, train/val/test splits, overfitting, classical algorithms, and evaluation metrics.
Neurons, activation functions, backpropagation and gradient descent, CNNs, and RNN/LSTM limitations.
Self-attention, positional encoding, multi-head attention, and the encoder-decoder vs. decoder-only distinction.
Scaling laws, pretraining, context windows, and the SFT/RLHF alignment stages that make a model helpful.
Zero/few-shot prompting, chain-of-thought, system vs. user prompts, sampling parameters, and prompt evaluation.
Tool/function calling, the ReAct loop, MCP, and agent memory.
Task decomposition, orchestrator/subagent topologies, and the LangGraph/CrewAI framework landscape.
Chunking, embeddings, retrieval and re-ranking, and why grounding reduces but never eliminates hallucination.
When to fine-tune vs. use RAG or prompting, LoRA/QLoRA, RLHF/DPO, and dataset quality.
GPUs/TPUs/NPUs, CUDA and the software stack, memory bandwidth, and edge AI constraints.
API vs. self-hosted, quantization, KV-caching, batching, and latency-vs-throughput tradeoffs.
The 8-stage AI lifecycle, the seven core AI engineering disciplines, and designing for AI-specific failure.
Hallucination, prompt injection, bias, guardrails, human-in-the-loop design, and the EU AI Act.
Computer vision inspection, CAD/BIM copilots, and AI for NEC calculations, protection coordination, and PV sizing.
No-code automation with n8n/Make, the AI Agent node, and when to graduate to custom code.
Real diagnostic scenarios — retrieval failures, rate-limit degradation, agent tool loops, and eval/production drift.
A grounded help-center chatbot — chunking strategy, vector-index sizing, and a retrieval + re-ranking pipeline.
An orchestrator + researcher subagents + writer agent — task decomposition and a worked cost/latency calculation.
LoRA fine-tuning vs. a general LLM call — a worked cost comparison and asymmetric precision/recall tradeoffs.
Self-hosted vs. API — a worked cost-crossover calculation and a latency/throughput budget.
A CV triage pipeline for inspection photos — precision/recall tradeoffs and a human-in-the-loop review workflow.
Requirements definition, AI-specific design reviews, model cards, vendor evaluation, and versioning discipline.
A downloadable kit of 12 real project templates, from a feasibility worksheet to a project handoff runbook.
A final internal-knowledge-assistant design assignment, drawing-interpretation and troubleshooting exercises, and your certificate of completion.
12 real project templates — a model & use-case feasibility worksheet, a prompt design & evaluation worksheet, a RAG pipeline design worksheet, a fine-tuning dataset & hyperparameter worksheet, a model evaluation & benchmark report template, an AI system architecture review checklist, an AI safety & risk assessment checklist, a deployment & inference cost worksheet, a monitoring & observability checklist, an AI debug & incident log template, a data governance & privacy checklist, and a project handoff & runbook template. Available for download from Module 20 (Documentation & Resource Kit) once you've unlocked the program.
Coming later: sample agent/RAG project code and video demonstrations. Not included in this release.