Chapter 88 — How to Use These Roadmaps
📖 Definition
A role roadmap is an ordered plan that answers three questions for one job title: what to learn, in what order, and what to build to prove it. It is not a syllabus. A syllabus lists topics; a roadmap ends every stage in a shippable artefact you can put in front of an interviewer.
🔍 Why This Section Exists
The rest of this book prepares you for the interview you already have. This section decides which interview to chase. In the current market the highest-demand titles are no longer "MERN developer" — they are AI engineer, ML engineer, data engineer, MLOps, and the newest hybrid, forward deployed engineer (FDE).
Your existing full-stack skills are not wasted. Four of the six roadmaps below start from exactly where a React/Node developer already stands.
Pick one role — the comparison table
| Role | Entry bar | Time to job-ready (part-time) | Best previous background | Chapter |
|---|---|---|---|---|
| AI Engineer | Any working developer | 6-9 months | Full-stack, backend | 89 |
| ML Engineer | Python + real maths | 12-18 months | Data analyst, Python backend, CS grad | 90 |
| Prompt Engineer | Precise writing + systematic testing | 3-5 months | QA, tech writer, domain expert, analyst | 91 |
| Forward Deployed Engineer | Ships end to end already | 9-12 months | Full-stack with client exposure | 92 |
| Data Engineer | SQL + Python | 6-10 months | Analyst, backend | 93 |
| MLOps / AI Platform | Linux, Docker, one cloud | 6-10 months | DevOps, SRE, infra | 94 |
If you are a MERN developer reading this and cannot decide: AI Engineer. It reuses your API, auth, database, and deployment skills, and the gap to fill is the smallest of the six.
🧱 Rules That Apply to Every Roadmap
- One roadmap at a time. Two half-finished paths read as neither on a CV.
- Ship in public. A deployed project beats a certificate in every one of these interviews. A GitHub repo with a live URL and a README that states the problem, the approach, and the numbers is the whole game.
- Measure everything you claim. "Faster", "better", and "more accurate" need a number beside them. This is the single biggest difference between a course-completer and a hire.
- Read the job ads before stage 1, and again at every stage. Requirements in AI move faster than any roadmap, this one included.
- Six focused hours a week finishes these. Twenty hours in one weekend a month does not.
- Do not pay for placement. No institute can sell you a job. Every tool in these roadmaps has a free tier or a free local equivalent.
📅 A Weekly Cadence That Works
| Slot | Hours | Activity |
|---|---|---|
| Two weekday evenings | 2 × 1.5 | Learn the current stage's topics — read, then re-type every code example by hand |
| One weekend block | 3 | Build the stage artefact; commit even when it is ugly |
| Friday, 30 minutes | 0.5 | Write down what broke this week and how you fixed it — this becomes your interview stories |
Six hours a week, one stage every 4-6 weeks, five stages: that is the arithmetic behind every timeline in this section.
🌍 How Employers Actually Read a Self-Taught CV
- The project section is read first, before education, before skills. Put two deployed projects at the top with a one-line result each.
- A number beats an adjective. "Cut p95 latency from 4.2s to 900ms with prompt caching" survives screening; "Worked on performance optimisation" does not.
- A written failure story raises trust. A README section titled "What went wrong and what I changed" signals production instinct.
- Vague AI claims lower trust. "Built an AI chatbot" invites the question "what did you actually own?" Be specific about the layer you built.
🎯 Likely Interview Questions
- Why are you moving from full-stack to AI/data engineering? — Point to the artefacts, not enthusiasm: what you built, what it does, what it measures.
- What have you built with no tutorial to follow? — Have one project where you chose the schema, the retries, and the deploy target yourself.
- How do you keep up with a field that changes this fast? — Name your actual loop: release notes, one paper or changelog a week, and a habit of re-baselining old prompts and models.
- This role needs production experience you do not have. Why should we hire you? — Answer with the closest real thing you own: something deployed, monitored, and used by someone other than you.
- Where do you want to be in two years? — Interviewers are checking that you picked a direction. A named specialism beats "anything in AI".
← AI: Cost & Latency Optimization | Index | Next: AI Engineer Roadmap →