75 days to the end of the year
Your full AI engineer roadmap as a daily sprint, Sunday October 18 to Thursday December 31, 2026. Fifteen real-world projects with pilot users, all 23 chapters of Hands-On LangChain, a ByteByteGo article every day, spaced revision, and something to present every week. Bonus work is marked in gold, for days when you have extra time. Plan on about 3 focused hours a day, plus 50 minutes on every seventh day.
Five phases
Each phase ends with a review day and a gate to pass before moving on.
The method
Ten rules that turn tutorials into skills and repos into CV evidence.
- One project at a time, 3β4 new skills maxSmall enough to finish, big enough to matter.
- Number targets before codeWrite the targets into the README first. "Recall@5 is 84%" beats "it works".
- Measured by a committed scriptEvery repo has an eval or bench script that prints the scoreboard.
- Each project feeds the nextOutputs become inputs, so the small projects add up to two capstones.
- Every core skill appears at least twiceThe second use is where it sticks. See the skill matrix.
- The book lands just before it's neededEach chapter is scheduled a few days before the project that applies it.
- One system design article every dayTen minutes of ByteByteGo matched to what you're building.
- Real users, real dataEvery project names a real person who uses it and real data it runs on.
- Revise on a schedule, present everythingTen minutes of spaced recall daily, and every project shown to a named audience.
- Time-box, then move onIf a target is still missed at the deadline, record the number and why, then continue.
A day in the sprint
On days without a book chapter, the 30 minutes go back to building. Python depth is on demand: when a project's Python tag matches the problem in front of you, open its card.
Day by day
Every day has a build task, a core article, and two bonus articles; book days add a chapter with a reading note. A day counts as complete when its build, book, core article, and revision are ticked, plus the weekly review and share on every seventh day.
Projects
Each project has a real-world brief, a scoreboard for your measured results, and a presentation for a named audience. P16 is a bonus project.
Revise, present, and illustrate
Learning sticks when you recall it on a schedule and explain it to someone else. Everything here is built into the day-by-day plan and earns XP.
Presentation kit
Tools for illustrating
For spaced revision, any flashcard app works; Anki is free and schedules reviews for you.
Capstones for your CV
Each capstone is assembled from finished small projects and tested by real pilot users. Fill every highlighted placeholder with a measured result only.
Real-world pilot plans
Nizam pilot: legal and compliance users
- Day 12 Message 3 potential pilot users: lawyers, compliance officers, or law students.
- Day 17 Collect at least 30 of their real questions for the eval set.
- Day 22 First live session: they try Nizam and rate 10 answers.
- Day 30 Phase show and tell with the pilot group.
- Days 53β62 They move to the live AWS URL.
- Day 71 15-minute usability sessions on the Next.js UI.
Raddad pilot: one real small business
- Day 25 Find one clinic, restaurant, or store willing to pilot.
- Day 36 Build the tools on their real FAQs, hours, and booking rules.
- Day 39 Turn their real customer questions into 30 test conversations.
- Day 41 Demo to the owner and collect a rating.
- Day 52 Phase show and tell with the owner.
- Day 58 Deploy for the business, with Langfuse tracking quality.
Anonymize customer data, get written permission before using anyone's messages, and never put real personal data in a public repo.
Skill coverage
Which project trains which skill. Dashed gold dots are the bonus project.
Your resource stack, rated
Relevance: how much the skill matters for AI engineer roles. Importance: how much your CV needs it now. Strength: resource quality at your level.
Python, when you need it
Not scheduled. When a project's Python tag matches the problem in front of you, open the card, learn just that, and get back to building.
Bonus Optional
Everything here is extra, for days when the core work is done and you have energy left. Each item earns XP and counts toward the Legend level.
Bonus learning
A Next.js admin dashboard for Raddad with a handoff inbox, Langfuse-linked conversations, analytics, and sign-in. It's in the Projects section.
Two per day in the day-by-day plan, chosen from the same theme as that day's work.
Chapters 8β9 for Nizam, and 13β14 and 18β20 for Raddad.