Choose the right AI and data starting point
Generative AI, machine learning, and data foundations overlap, but they do not require the same starting skills. Generative AI can provide quick exposure to prompting, evaluation, retrieval, and application design. Machine learning adds a stronger need for data preparation, feature thinking, metrics, and experimentation. Data science goes further into statistics, modelling choices, and communicating uncertainty.
Choose the first layer that helps you build and explain something correctly. Starting with foundations is not a delay when it prevents you from copying code you cannot diagnose or presenting model output you cannot evaluate.
- Start with data analytics when you need confidence with datasets, queries, visualisation, and business questions.
- Start with Generative AI when you can already code basic applications and want to build evaluated AI-assisted workflows.
- Start with machine-learning concepts when you are comfortable with Python, data manipulation, and basic quantitative reasoning.
- Consider deeper neural-network work after you can explain training data, validation, error analysis, and model trade-offs.
Check your foundations honestly
Readiness is not measured by the number of tools you recognise. Test whether you can load and inspect a dataset, write a small program, explain an average and a distribution, and distinguish training evidence from a claim about real-world performance. These foundations make advanced lessons easier to question and apply.
Also check your evaluation habits. AI projects need more than a working demo: you should define what a good answer looks like, collect representative test cases, review failures, and explain limitations. This is especially important when a generated response or prediction could influence another person.
- Can you transform and validate tabular data without hiding missing or inconsistent values?
- Can you break a programming problem into functions and inspect an error message methodically?
- Can you explain why accuracy alone may be a weak evaluation measure?
- Can you document where an AI system should defer to a person or trusted source?
Define a portfolio-sized proof project
A strong starter project has a narrow user, a clear input and output, and an evaluation plan. Examples include an analysed public dataset with a decision-focused dashboard, a retrieval-based assistant tested against a curated question set, or a small prediction workflow with transparent baseline comparisons.
Keep an evidence log as you build. Record the data source, assumptions, failed approaches, evaluation results, and the reason for each major decision. That record becomes more valuable in an interview than a polished interface with no explanation of how the system was tested.
- State the problem and who benefits from the result.
- Define a baseline before adding a more complex method.
- Keep private or sensitive data out of a public portfolio.
- Prepare a three-minute explanation covering the approach, evidence, limitations, and next improvement.
Key takeaways
Use this before choosing a cohort
Generative AI is a strong entry point when you need fast practical exposure.
Machine learning needs comfort with data, logic, and evaluation.
Deep learning should be approached with enough math and project patience.