← all writing career · 14-weekend roadmap

How to become an AI/ML Product Owner.

by Josen Joy 19 May 2026 9 min read free resources only

I am not a data scientist. I do not have a machine learning degree. But I want to own AI products and that is entirely possible in 2026, because the skills an AI Product Owner needs are not about building models from scratch. They are about understanding the landscape well enough to lead the teams that do.

This is the exact plan I am following. Every weekend is mapped, every resource is linked, and everything is free (or nearly free). If you are starting from zero with a bit of HTML or PHP experience, this is for you.

You do not need to become an engineer. You need to become fluent enough to lead engineers, evaluate their work, and define what success looks like.

What does an AI Product Owner actually do?

Before we start learning, it helps to understand the job. An AI PO sits between business stakeholders and the technical team. You translate problems into AI-shaped solutions. You define what "good" looks like in AI, that means writing evaluations. You prioritise features, manage expectations, and make sure the product ships without hallucinating or breaking something important.

The skills break into six categories. Think of this as the map before we dive into the territory.

01
AI/ML foundations
How LLMs work, what tokens are, why hallucinations happen, where the bodies are buried.
02
The modern AI stack
RAG, agents, vector databases, evals the tools powering AI products today.
03
AI product skills
Writing AI user stories, defining evals, managing AI failure modes, scoping safely.
04
Technical literacy
Reading Python, understanding APIs, prompting well. Not coding from scratch.
05
Hands-on demos
Three portfolio projects that prove you can think like a builder.
06
Responsible AI
Bias, fairness, privacy, and how to communicate uncertainty to non-technical stakeholders.

Phase 1 - Foundations

Before building anything, you need the mental model. These first weekends are all theory, but they are the fastest-moving theory you will ever watch. No code required. Just watch, take notes, and let the concepts settle.

W1
phase 1 · foundations
How LLMs actually work

Start here. Andrej Karpathy a co-founder of OpenAI and former head of AI at Tesla released a 3.5-hour deep dive into how models like ChatGPT are built from the ground up. It covers training, tokens, hallucinations, and why models behave the way they do. For a complete beginner, this is the single best video to watch first.

Also watch his shorter introductory video explaining LLMs in plain language. Together, these two establish everything else in this roadmap.

W2
phase 1 · foundations
Generative AI for Beginners (Microsoft)

Microsoft's free 18-lesson course on GitHub covers everything from "what is a large language model" to building your first GenAI application. Each lesson has written content and a video. Work through lessons 1–6 this weekend: introduction to GenAI, LLMs, responsible AI, prompt engineering, and advanced prompting.

W3
phase 1 · foundations
Prompt engineering + DeepLearning.AI short courses

DeepLearning.AI (Andrew Ng's platform) offers free short courses averaging 1–2 hours each. Start with ChatGPT Prompt Engineering for Developers do not be put off by the title, it is beginner-friendly and teaches the skill every AI PO needs most. Then do Building Systems with the ChatGPT API.

W4
phase 1 · foundations
How to evaluate AI the most underrated skill

Evals are how you define "done" in AI products. Watch Krish Naik's introduction to LLM evaluation on YouTube. Then read the LangSmith evaluation documentation (no code needed just understand the concepts: accuracy, relevance, groundedness, latency). This sets you up for your first portfolio project.

By the end of this weekend, you should be able to write an evaluation framework for a simple chatbot on paper, without any code.

§ project 1

Portfolio project 1 eval framework writeup. After Week 4, you have enough knowledge to write your first portfolio piece. Pick a simple AI use case (a customer-service chatbot, a document summariser, a FAQ bot). Define the evaluation framework: which metrics, which test cases, what does pass/fail look like? Write it up as an article on your website. No code required this is a thinking and communication exercise, and it is more useful than another tutorial demo.

Phase 2 - The modern AI stack

Now we go deeper into the tools that power real AI products. This phase introduces Python at a reading level (you do not need to write it from scratch), and covers the two most important architectural patterns in enterprise AI today: RAG and agents.

W5
phase 2 · the stack
Python basics for AI POs (reading level)

You do not need to code Python from memory. You need to read it and understand what is happening. Watch freeCodeCamp's Python for Beginners first two hours only. Focus on: variables, functions, lists, dictionaries, and calling APIs. After this weekend, you should be able to look at a Python script that calls OpenAI and explain what each section does.

W6
phase 2 · the stack
What is RAG and why does every AI product use it?

RAG (Retrieval-Augmented Generation) is how AI products access your company's data without retraining a model. Watch Krish Naik's "Complete RAG Crash Course with LangChain in 2 Hours" hands-on, beginner-friendly, covering everything from document loading to vector search to generating answers.

You do not need to run the code yourself this weekend. Watch to understand the concepts: embeddings, vector databases, chunking, retrieval, generation.

W7
phase 2 · the stack
Build your first RAG demo

Now run the code. Follow along with Krish Naik's tutorial using Google Colab (free, runs in your browser no local Python setup needed). You will build a simple document Q&A system. Your goal is not to write the code from scratch, but to follow along, understand each step, and be able to explain the full pipeline.

Stuck on something? Use Claude to explain any line you do not understand. This is legitimate learning, not cheating.

W8
phase 2 · the stack
Write up your RAG demo as an article

Document what you built, how it works, what you would do differently, and what an AI PO needs to know when scoping a RAG feature. Screenshots, diagrams, honest reflections those make this more valuable than a polished tutorial. Publish it on your site. This is portfolio project 2.

Writing about what you built cements the learning and builds your public profile at the same time. Two birds, one stone.

Phase 3 - Agents and agentic AI

This is the frontier. Agents are AI systems that do not just answer questions they take sequences of actions, use tools, and complete goals. LangGraph is the most widely used framework for building them. Companies like Klarna, Replit, and Elastic are already using it in production.

W9
phase 3 · agents
What are AI agents and why do they matter?

Watch DeepLearning.AI's free AI Agents in LangGraph course (available on Coursera, auditable for free). It is co-taught by the founder of LangChain and covers the core concepts: what an agent is, how it uses tools, how state flows through a graph, and what human-in-the-loop means.

Also watch the freeCodeCamp "Build Powerful AI Agents with LangGraph" beginner masterclass on YouTube. Watch only, do not code yet.

W10
phase 3 · agents
Understanding LangGraph's building blocks

Watch the freeCodeCamp LangGraph Complete Course for Beginners. This covers nodes, edges, state, and the graph pattern in depth. After this weekend you will understand how an agent decides what to do next, how tools get called, how state is tracked, and how the graph loops until a task is complete.

This is also where knowledge files (.md files) come in a way to give your agent context about what it needs to know to do its job.

W11
phase 3 · agents
Build a simple LangGraph agent (follow-along)

Follow Krish Naik's Getting Started with Agentic RAG with LangGraph tutorial. Build a simple research agent that can search, retrieve, and summarise. Use Google Colab. Your goal is to run a working agent and understand every part of the graph.

LangGraph runs in Python. If you are not comfortable, paste the error into Claude and ask what it means.

W12
phase 3 · agents
Write up your agentic AI demo

Document your LangGraph agent. Explain: what problem it solves, how the graph is structured, what tools it uses, what you would do differently in a production product, and what an AI PO needs to know when scoping an agentic feature. Publish it. This is portfolio project 3.

Phase 4 — Responsible AI & portfolio

W13
phase 4 · responsible AI
Bias, fairness, privacy, safety

Take Google's free Responsible AI course on Coursera (roughly 8 hours, auditable free). It covers hallucinations and how to mitigate them, bias in training data, privacy considerations, explainability, and frameworks for AI governance. As an AI PO, this is the material that protects you and your company.

W14
phase 4 · portfolio
Build your portfolio page and start applying

Create a dedicated AI PO portfolio page on your site. Link your three projects. Write a short summary of what you have learned and what problems you want to solve as an AI PO. This is not a CV it is evidence of thinking and doing. Start applying for junior AI PO, Associate Product Manager (AI), or AI Product Analyst roles.

The full plan, at a glance

WeekendTopicKey resourceOutput
W1How LLMs workAndrej KarpathyMental model
W2GenAI foundationsMicrosoft GenAI for BeginnersNotes
W3Prompt engineeringDeepLearning.AI short coursesPrompts
W4EvaluationsKrish Naik + LangSmith docsEval writeup draft
W5Python for readingfreeCodeCampCan read Python
W6RAG conceptsKrish Naik RAG crash courseUnderstand the stack
W7Build RAG demoGoogle Colab + tutorialWorking demo
W8RAG writeupPortfolio project 2
W9Agent conceptsDeepLearning.AI + freeCodeCampUnderstand agents
W10LangGraph deep divefreeCodeCamp full courseUnderstand the graph
W11Build agent demoKrish Naik agentic RAGWorking agent
W12Agent writeupPortfolio project 3
W13Responsible AIGoogle on CourseraFramework knowledge
W14Portfolio + applyReady to apply
§ a note on PHP

Most of these tutorials use Python. You do not need to rewrite them in PHP. Your job as a PO is to understand what is being built, not replicate it. Follow along in Python using Google Colab (free, runs in your browser) no installation needed. The concepts transfer directly to any language or platform you later work with.

What to do this weekend

Open YouTube, find Andrej Karpathy's channel, and watch his "Deep Dive into LLMs like ChatGPT". Make notes. Come back next weekend for Microsoft's GenAI for Beginners. That is it. One video, one weekend, one step.

The AI PO role is genuinely reachable from where you are. Not in years in months. The field is moving fast, which means the bar for "experienced" is still low enough that consistent effort over fourteen weekends puts you ahead of most applicants.

The goal is not to become a data scientist. It is to become the person data scientists trust to point them at the right problem.

I will be updating this article as I work through the plan. If you are following along, drop me a line on LinkedIn or hello@ainika.xyz and let me know how it is going.