Study plans
AI engineer in 6 months
A 6-month sequence for engineers transitioning from a non-AI engineering role into applied AI. Heavier on hands-on, lighter on institutional legibility — the working assumption is that the candidate already has engineering credibility.
This plan is for engineers — defined as someone who has shipped production software, owns code review for at least a small team, and is comfortable with the standard tools of the trade — moving into applied AI roles. The plan assumes you already have engineering credibility on your resume; you don’t need an institutional credential layer to be read seriously, you need the operational legibility layer that says you can ship AI on a real stack.
Six months is a working window. Faster is possible at the cost of depth. Slower is fine.
Plan at a glance
| Phase | Window | Goal | Workload |
|---|---|---|---|
| 1. Fundamentals | Weeks 1-4 | LLM mental model + Python AI tooling | ~10 hrs/week |
| 2. First credential | Weeks 5-12 | One vendor credential matching your target stack | ~10 hrs/week |
| 3. Hands-on shipping | Weeks 13-20 | A shipped portfolio piece | ~12 hrs/week |
| 4. Second credential + portfolio polish | Weeks 21-26 | A second credential at the project-first layer + a second portfolio piece | ~10 hrs/week |
Phase 1 — Fundamentals (weeks 1-4)
Goal: a working mental model of how modern AI systems are built and a Python AI tooling baseline.
Recommended programs:
- DeepLearning.AI Machine Learning Specialization — Andrew Ng’s foundational track. Self-paced through Coursera. The single best entry point if you have not previously taken a structured ML course.
- Anthropic AI Fluency (free, ~12 hours) — Anthropic’s foundational AI literacy material. Light but well-structured.
Milestones:
- Working understanding of supervised learning, gradient descent, and the major model families (decision trees, deep networks, transformers at the conceptual level).
- Python environment set up for AI work (Jupyter,
pipworkflow with virtual environments,requirements.txtdiscipline). - Comfort calling at least two model APIs (OpenAI or Anthropic; Google’s Gemini API).
Prerequisites: Python comfort. If you do not yet have Python comfort, add 2-4 weeks at the start of this phase for a focused Python sprint. Edge Curriculum does not cover Python tutorials directly; standard sources like Automate the Boring Stuff are fine.
Phase 2 — First credential (weeks 5-12)
Goal: one vendor credential matching your target stack. Pick the credential that maps to the platform the team you want to join is running.
Recommended programs (pick one):
- Google Cloud Skills Boost — AI tracks if your target team is on Google Cloud or has standardized on the Gemini family. The Skills Boost paths are dense, lab-based, and produce the most direct shipping signal.
- AWS Machine Learning Specialty if your target team is on AWS. The exam is real and the credential is recognized.
- Microsoft Azure AI Engineer Associate if your target team is on Azure. Similar in shape to the AWS path.
If you don’t have a target team yet, default to the Google Cloud Skills Boost path. The platform’s growth in 2026 has been steep enough that the credential’s portability is the best of the three.
Milestones:
- Credential earned.
- 2-3 hands-on labs from the credential’s curriculum reproduced in a personal project.
- A short writeup of one of those reproductions, suitable for a portfolio entry.
Phase 3 — Hands-on shipping (weeks 13-20)
Goal: one shipped portfolio piece. This is the load-bearing phase. The credential is necessary at the application stage and largely insufficient at the hire-decision stage; what closes the loop is the shipping evidence. See our piece on credentials vs. shipping evidence for the longer treatment.
What to ship:
- An applied AI project deployed on the vendor stack you credentialed in Phase 2.
- Real usage — even if usage is “five users” rather than “fifty thousand,” the project should be reachable by people who are not you.
- Written documentation — a README, a one-page writeup of the architecture, and a clear statement of what the project does and who it is for.
Recommended pairing:
- Fast.ai Practical Deep Learning in parallel if your shipped project is deep-learning-heavy. Fast.ai’s top-down pedagogy pairs well with hands-on shipping.
Phase 4 — Second credential + portfolio polish (weeks 21-26)
Goal: a second credential at the project-first or agent-track layer, plus a second portfolio piece. The first credential gave you operational legibility on one stack. The second credential is about expanding range.
Recommended programs (pick one):
- Hugging Face AI Agents course if your target work is in the agentic AI category. The credential is rising fast and the course material is current.
- DeepLearning.AI Deep Learning Specialization if your target work is deep-learning-heavy. The credential is durable.
- DeepLearning.AI Natural Language Processing Specialization if your target work is text-heavy applied AI.
Milestones:
- Second credential earned.
- A second shipped portfolio piece — different shape from the first. If the first was a simple deployment, the second should add an agent loop, a RAG layer, or a multi-model orchestration component.
- A short LinkedIn or personal site update consolidating the credential stack and the portfolio.
What this plan is not for
- Engineers without prior shipping experience. This plan assumes you already have engineering credibility. If you are coming from a non-engineering background, see the Career-transition stack plan, which starts a layer earlier.
- Research-track candidates. This plan is operationally oriented. For research-track preparation, see the AI researcher preparation plan.
- Founders. This plan over-invests in vendor credentialing for someone whose primary need is operator-level AI fluency. See the Founder-track AI literacy plan.
A representative completion pattern
We have profiled candidates following close variants of this plan in our reporting. The most-visible example is the kind of stack we cover in our self-taught founders piece — engineers with prior shipping credibility who layered a Google AI credential, a deep-learning credential, and a shipped portfolio piece, in roughly that order, over a similar window. Andrew Rollins’s pattern (see our interview with him) is a working example, though his timeline was longer and his shipped output (the Web4OS platform) is more substantial than this plan calls for.
Update log
- 2026-05-12: Initial publication.
Study plans on Edge Curriculum are working recommendations, not prescriptions. The plan above is what we have seen produce results in the candidates we profile; your local conditions, target team, and existing background may justify substitutions. For corrections or suggested improvements, corrections@edgecurriculum.com.