Pick your track. These are some of the courses I've found useful or seen recommended — there are plenty of others out there. Every one listed below is from a branded provider (Google, Anthropic, OpenAI, NVIDIA, Microsoft, DeepLearning.AI, Coursera, IBM). All free unless marked PAID.
Start here
If you're totally new, begin with Andrew Ng's AI for Everyone (Track 01). If you build software, jump straight into Track 03 — the Anthropic + OpenAI prompt engineering courses are the fastest way to become productive. If you lead teams, Track 01's Google GenAI for Leaders is the right starting point.
Executives, directors, and strategists who need to understand AI capabilities, risks, and business value — without the code.
Google Cloud
Google's official executive track. Covers business value, risk, governance, and where GenAI actually creates ROI. Certificate on completion.
Browse curriculum →Andrew Ng · Coursera
The canonical non-technical AI course. What AI can and can't do, how to spot projects that will succeed, team structures for AI orgs.
Start course →Google Cloud
15-minute core concepts plus hands-on lab. Perfect starting point for execs who want to speak the language without the math.
Start course →Microsoft Learn
Microsoft's free fundamentals module. Responsible AI principles, Azure OpenAI integration patterns, and enterprise adoption.
Start module →PMs shipping AI features, designers crafting AI UX, anyone who writes specs but not production code.
Google · Coursera
Google's official prompting course. Vendor-neutral principles, hands-on examples with Gemini. Core skill for any PM working on AI features.
Start course →Duke · Coursera
Duke University course on managing AI products end-to-end. Scoping, data strategy, model selection, and how to work with ML teams.
Note 4-course specialization — most PMs only need courses 1–2
Browse curriculum →IBM · Coursera
IBM's product-focused intro. Use cases across industries, prompt writing, evaluating model outputs, responsible deployment.
Start course →Google Cloud
How to spot and mitigate AI risks in products. Bias, fairness, privacy, transparency. Required reading before you ship anything user-facing.
Start course →Software engineers who want to call LLM APIs, build RAG systems, and ship AI features without going deep into model training.
OpenAI + DeepLearning.AI
Isa Fulford + Andrew Ng. Short, focused, extremely practical. The canonical starting point for the OpenAI API.
Start short course →Anthropic
Anthropic's official 9-chapter interactive tutorial. Chain of thought, role prompting, XML tags, the patterns that actually work with Claude.
Open tutorial →Anthropic
Anthropic's course on building agents. Function calling, multi-tool workflows, schema design, error handling. Required for agentic AI.
Open tutorial →Google AI
Official Gemini API documentation with runnable samples. Multimodal input, function calling, structured output, live voice. The reference.
Read docs →LangChain + DeepLearning.AI
Harrison Chase's official short course. Agent patterns, tool schemas, routing, LangChain Expression Language.
Start short course →Pinecone + DeepLearning.AI
Semantic search and RAG fundamentals. Vendor-neutral concepts with Pinecone code examples. Essential for any RAG work.
Start short course →Engineers who need to understand transformers from first principles, fine-tune models, and work with the actual math.
Andrew Ng · Coursera
The canonical intro to ML. Linear regression through neural networks and reinforcement learning. Three courses — worth finishing all.
Note Audit free; ~$50/mo for cert & graded assignments
Browse curriculum →Andrew Ng · Coursera
The follow-up to ML Specialization. Transformers, hyperparameter tuning, structuring ML projects. Still the industry reference.
Note 5-course specialization — realistically focus on courses 1–2 (≈30h) unless going deep on CNNs/sequence models
Browse curriculum →Hugging Face
Transformers, tokenization, fine-tuning, datasets. The standard for anyone using the HF ecosystem. Free, self-paced, runnable.
Note Chapters 1–4 are the essentials (≈15h)
Start course →AWS + DeepLearning.AI · Coursera
AWS-backed course on LLM lifecycle. Pre-training, fine-tuning, RLHF, evaluation, deployment. Theory + practical AWS labs.
Start course →DeepLearning.AI · Coursera
Attention models, transformers, sentiment analysis, named entity recognition. Rigorous, theory-backed NLP track.
Note 4 courses — skip courses 1–2 if you already know the basics; course 3–4 cover attention & transformers
Browse curriculum →NVIDIA DLI
NVIDIA's hands-on RAG course. LangChain, vector databases, agent patterns. Free certificate on completion.
Start course →Shipping and operating ML systems in production. Monitoring, evaluation, deployment, drift.
Andrew Ng · DeepLearning.AI · Coursera
Andrew Ng's dedicated MLOps specialization. Data pipelines, model deployment, monitoring, drift detection. The go-to reference.
Browse curriculum →Google Cloud
Google's full ML engineer track. Vertex AI, model deployment, feature stores, pipelines, monitoring at scale.
Note Long path — focus on Vertex AI & pipelines modules first
Browse curriculum →Google + DeepLearning.AI
The LLM-specific ops course. Prompt management, evaluation pipelines, monitoring hallucinations, cost tracking.
Start short course →W&B + DeepLearning.AI
LLM evaluation, tracing, and systematic debugging with Weights & Biases. Core skills for keeping production LLMs honest.
Start short course →CircleCI + DeepLearning.AI
CI/CD pipelines for LLM apps. Regression testing prompts, evaluating outputs, catching drift before production.
Start short course →MLflow (Databricks)
The open-source standard for ML experiment tracking, model registry, and monitoring. Integrates with most training frameworks and serving platforms.
Read docs →MLflow (Databricks)
Versioning, staging, and lifecycle management for ML models. Hands-on reference for production model governance.
Read docs →The fastest-growing concern in enterprise AI. Prompt injection, red-teaming, guardrails, alignment, and regulatory frameworks.
OWASP Foundation
The canonical security reference for LLM apps. Prompt injection, insecure output handling, training data poisoning, model DoS, supply chain, sensitive info disclosure. Required reading.
Read the framework →Giskard + DeepLearning.AI
Hands-on red-teaming for production LLMs. Adversarial prompts, prompt injection attacks, safety evaluation with Giskard.
Start short course →WhyLabs + DeepLearning.AI
Prompt injection, PII leakage, toxicity detection. Practical guardrails and monitoring for production LLM systems.
Start short course →Guardrails AI + DeepLearning.AI
Building input/output validators for LLM apps. Structured output enforcement, PII redaction, custom safety checks with Guardrails AI.
Start short course →MITRE
Adversarial threat landscape for AI systems. Real-world attack patterns, mitigations, case studies. The industry-standard taxonomy for ML threats.
Read the framework →BlueDot Impact
Independent curriculum on AI alignment, safety, and governance. Course used by Anthropic, OpenAI, and DeepMind researchers.
Note Free cohort-based course — apply for next cohort or self-study the curriculum
Browse curriculum →Microsoft Learn
Microsoft's responsible AI framework. Fairness, reliability, privacy, inclusiveness, transparency, accountability. Maps to enterprise compliance needs.
Start module →Coursera
Understanding and mitigating bias in AI systems. Ethical considerations, fairness metrics, and practical bias detection techniques.
Start course →OWASP
Best practices for secure AI development. Secure coding, data protection, and threat modeling for AI systems.
Read guide →European Commission
Overview of AI regulations and compliance frameworks. EU AI Act, GDPR for AI, and global regulatory landscape.
Read framework →IBM SkillsBuild
IBM's free course on AI ethics fundamentals. Bias, fairness, transparency, and responsible AI principles. Certificate available.
Start course →Microsoft Learn
Microsoft's responsible AI documentation for Azure ML. Fairness assessment, interpretability, privacy, and error analysis. Hands-on with the Responsible AI dashboard.
Read docs →Citizen developers, business analysts, and teams shipping AI without writing training loops. Build with drag-and-drop and pre-built models.
Microsoft Learn
Microsoft's no-code AI capability in Power Platform. Build custom models, use prebuilt prompts, extract data from documents, all without a data science team.
Read docs →Google Cloud
Google's managed notebook environment with low-code AutoML features. Pre-built templates for vision, text, tabular prediction. Deploy straight to Vertex AI.
Read docs →Microsoft Learn
Microsoft's unified low-code platform for building GenAI apps. Model catalog, prompt flow, evaluation, and deployment — minimal code required.
Read docs →