AI-Adjacent Roles Bootcamp — Program
AI-Adjacent Roles Bootcamp — Program
A paced, ~10-week program that turns the AI-adjacent path from a self-serve course into a bootcamp: it sequences the lessons (Topics 1–10), the capstone deliverables (real artifacts, AI-reviewed against rubrics), the interview drills, and career services into one journey with weekly milestones. The whole bootcamp orbits one project — the FjordPay Support Copilot — so that by Week 10 every artifact you've built tells a single coherent story you can pitch in an interview.
It's self-paced — the weeks are a recommended rhythm (roughly 5–8 focused hours each), not hard deadlines. You meet your capstone scenario in Week 1 and carry the same feature through every deliverable.
Core program (Weeks 1–10)
| Week | Learn (topics) | Capstone deliverable | Interview practice | ~Hours |
|---|---|---|---|---|
| 1 | Topic 1 (What are AI-adjacent roles?), Topic 2 (The AI roles map) — (+ general Core Tech Baseline alongside) | Meet the capstone — read the FjordPay Support Copilot brief; pick the AI-adjacent lens you'll own (AI PM / prompt engineer / AI-operations) | Behavioral & Curiosity: start drafting your STAR stories and "why AI" narrative | 5 |
| 2 | Topic 3 (How modern AI actually works) | (scope your Copilot — sketch where AI helps vs. where it must not act alone) | Behavioral & Curiosity (D1–D3) | 6 |
| 3 | Topic 4 (Prompt engineering fundamentals) | D1 — Prompt spec & iteration log | Practical Prompt-Craft (D1–D3) | 7 |
| 4 | Topic 5 (Evaluating AI output and quality) | D2 — AI output evaluation & rubric | AI Output Evaluation (D1–D3) | 8 |
| 5 | Topic 6 (Data, training, and the human-in-the-loop) | (revise D1–D2 from feedback; gather PII/edge-case notes for the risk register) | Practical Prompt-Craft (D4–D6) + Mock-interview checkpoint #1 (prompt-craft + output-evaluation) | 6 |
| 6 | Topic 7 (AI limitations, risks, and ethics) | D3 — Responsible-AI risk register & human-in-the-loop plan | Judgment & Ethics (D1–D3) | 7 |
| 7 | Topic 8 (AI Product Management) | D4 — AI feature PRD | Domain-AI Scenario (D1–D3) | 8 |
| 8 | Topic 9 (Tools and the AI ecosystem) | D5 — Hands-on build write-up (with RAG) | Domain-AI Scenario (D4–D6) + Judgment & Ethics (D4–D6) | 8 |
| 9 | Topic 10 (Landing an AI-adjacent role) + general Employability & Job Hunting | D6 — Portfolio case study & pitch | Behavioral & Curiosity (D4–D6) | 7 |
| 10 | (review & integrate — no new topic) + general Employability & Job Hunting | Assemble the portfolio (all 6 deliverables into one narrative) | Mock-interview checkpoint #2 (full mock across all five drill types) + career services (below) | 8 |
By the end of Week 10 you have: the full curriculum learned, a 6-artifact portfolio for one AI feature (prompt spec, graded eval, risk register, AI PRD, hands-on build, case study + pitch), repeated interview practice across all five drill banks, and your resume/LinkedIn ready.
The capstone thread
Every deliverable is the same feature at a later stage, so the schedule is built so each one unlocks right after the topic that teaches it:
| Deliverable | Title | Unlocks after | Due (week) | ~Minutes |
|---|---|---|---|---|
| D1 | Prompt spec & iteration log | Topic 4 | 3 | 75 |
| D2 | AI output evaluation & rubric | Topic 5 | 4 | 90 |
| D3 | Responsible-AI risk register & human-in-the-loop plan | Topic 7 | 6 | 75 |
| D4 | AI feature PRD | Topic 8 | 7 | 90 |
| D5 | Hands-on build write-up (with RAG) | Topic 9 | 8 | 90 |
| D6 | Portfolio case study & pitch | Topic 10 | 9 | 60 |
The capstone scenario (Ship an AI Feature End-to-End: The Support Copilot) is described in capstone/capstone.md; each deliverable lives in capstone/deliverables/ with its own instructions, worked example, and rubric, and starts from a template in templates/.
Interview drill banks
Practice is spread across all five banks so no skill is rehearsed only once:
- Practical Prompt-Craft (
interview/practical-prompt-craft.md) — live prompt-writing and prompt-critique (Topic 4). Weeks 3 & 5. - AI Output Evaluation (
interview/ai-output-evaluation.md) — output-critique and quality drills (Topic 5). Week 4 + checkpoint #1. - Judgment & Ethics (
interview/judgment-and-ethics.md) — trust, hallucination/bias/privacy, human-in-the-loop, disclosure (Topics 6–7). Weeks 6 & 8. - Domain-AI Scenario (
interview/domain-ai-scenario.md) — applied AI-PM scenarios: "how would AI help in [domain] and what would you watch out for?" (Topics 8–9). Weeks 7 & 8. - Behavioral & Curiosity (
interview/behavioral-and-curiosity.md) — STAR stories, catching an AI error, staying current, why you're pivoting (Topic 10). Weeks 1–2 & 9.
Career services (lite)
Delivered mainly in Weeks 9–10, drawing on the general Employability & Job Hunting track (content/general — Topic 07 LinkedIn profile, Topic 09 Resumes and applications, Topic 10 Interviewing and landing the offer). These are self/AI-reviewed against rubrics (no human mentor required).
Resume review rubric
- Outcome-focused bullets — 1: duties only · 2: some results · 3: quantified outcomes ("cut draft-reply time from 14h to X").
- AI-adjacent vocabulary & relevance — 1: generic · 2: some AI terms · 3: clearly reframed for AI-adjacent work (prompting, evaluation/rubrics, human-in-the-loop, RAG, responsible use).
- Portfolio linked — 1: none · 2: mentioned · 3: linked Support Copilot case study, tailored to the role.
- Clarity & format — 1: cluttered/typos · 2: clean · 3: crisp, one page, scannable in seconds.
Worked example — reframing one bullet (FjordPay Support Copilot). Take a real line off your current resume and rebuild it against all four criteria. Here a generic support-ops duty becomes an AI-adjacent, outcome-focused bullet:
Before (1/3 — duty only, no AI signal): "Responded to customer support tickets and escalated complex issues to senior staff."
After (3/3 — outcome + AI-adjacent vocabulary + portfolio hook): "Designed and evaluated the FjordPay Support Copilot, an LLM draft-reply assistant for payment-dispute tickets: wrote the prompt spec, built a 5-criteria eval rubric scoring 60 sampled outputs (raised pass rate 68% → 91% over three iterations), and added a human-in-the-loop gate so the AI never auto-sends refunds. Cut median first-draft time from ~14 min to ~3 min. [Case study: link]"
Why it scores 3 on each: it leads with quantified outcomes (68% → 91%, 14 → 3 min), uses real AI-adjacent vocabulary (prompt spec, eval rubric, human-in-the-loop) instead of buzzwords, and links the capstone case study. Keep it to 2–3 lines so it stays scannable. Note what is not claimed: you didn't "build an LLM" — you specced, evaluated, and de-risked one, which is exactly the AI-adjacent story. Run your own bullets through this same before → after pass before the AI review.
LinkedIn review rubric
- Headline signals target role — 1: old title only · 2: mentions AI · 3: clear target + value ("Aspiring AI PM / Prompt Engineer | ex-[field]").
- About tells the pivot story — 1: empty/generic · 2: present · 3: confident pivot story (where heading, what you bring from your domain, why AI, proof).
- Skills & proof — 1: none · 2: some · 3: AI-adjacent skills listed (prompt design, eval, responsible AI) + the capstone case study shown.
- Activity — 1: inactive · 2: occasional · 3: connecting/engaging in the AI field.
Worked example — headline + About paragraph. The headline names the target role and the value you carry over; the About paragraph runs the full pivot arc — where heading → what you bring → why AI → proof. Sample (assuming a customer-support background pivoting to AI PM):
Before headline (1/3): "Customer Support Team Lead at Acme Retail" After headline (3/3): "Aspiring AI Product Manager / Prompt Engineer | ex-Customer Support Lead | I turn LLM features into shippable, evaluated products"
About (3/3): "I'm moving into AI product and prompt engineering — building AI features that are actually safe to ship, not just demos. Six years leading customer-support teams taught me where automation breaks: edge cases, angry customers, and the moments a wrong answer costs trust. I bring that judgment to AI work.
Most recently I designed the FjordPay Support Copilot, an LLM assistant that drafts replies for payment-dispute tickets. I wrote the prompt spec, built an evaluation rubric and graded 60 real outputs against it (lifting the pass rate from 68% to 91%), wrote a responsible-AI risk register, and designed a human-in-the-loop gate so the model never auto-sends a refund. The result cut first-draft time from ~14 minutes to ~3.
What I'm looking for: an AI-adjacent role — AI PM, prompt engineer, or AI operations — where domain judgment and evaluation discipline matter as much as the model. Case study and portfolio linked below. Always happy to compare notes on prompt evals and human-in-the-loop design."
Why it scores 3: the headline states a concrete target plus a value proposition (not just "interested in AI"); the About moves through every beat of the pivot story, names AI-adjacent skills (prompt spec, eval rubric, responsible-AI, human-in-the-loop) and backs them with the same quantified proof as the resume, and closes with an invitation that doubles as an Activity prompt. Swap in your own domain and numbers, then run it through the AI review.
Mock-interview checkpoints
Use the AI mock interview: Checkpoint #1 (Week 5) — practical prompt-craft + AI output evaluation; Checkpoint #2 (Week 10) — a full mock across all five drill banks (prompt-craft, output-evaluation, judgment-and-ethics, domain-AI-scenario, behavioral-and-curiosity). Track which rubric criteria you repeatedly miss and focus there.
Salary-negotiation practice
A short drill (Week 10, grounded in general Topic 10 Interviewing and landing the offer): given a sample offer for an AI-adjacent role, practice (1) asking for time to consider, (2) researching the market range for the title and region, (3) making one polite counter tied to your value — including the de-risking and evaluation work in your portfolio — and (4) evaluating fit/growth in a fast-moving field, not just pay. The AI can role-play the recruiter and give feedback.
Worked example — a filled-in counter. Walk through the four steps on a concrete offer so you have a model to imitate, not just a checklist.
Sample offer: Associate AI Product Manager, $98,000 base + standard benefits. Recruiter: "We'd love to have you — can you let me know by Friday?"
Step 1 — buy time (don't accept on the call): "Thank you, I'm genuinely excited about this — the Copilot work is exactly the direction I want to grow. I'd like to review the full package and get back to you by Thursday. Does that work?"
Step 2 — research the range: check Levels.fyi, Glassdoor, and a recent local salary survey for "AI PM / associate PM" in your city/remote band. Suppose the median for the title and region lands around $108k, with the band running ~$100k–$118k. That puts the offer below median — a fair, evidence-based reason to counter.
Step 3 — one polite counter tied to value (the script): "I've done some research on the market for associate AI PM roles in this region, and the range I'm seeing centers around $108–112k. Based on that, and on what I'd bring — I've already shipped an evaluated, de-risked LLM feature end to end, including the prompt spec, the eval rubric, and the human-in-the-loop design that keeps it safe — would you be able to move the base to $110,000? I'm flexible on how we get there if part of it is a sign-on or an early review."
Step 4 — weigh more than pay: before you accept any number, score the role on growth signals — will you own real AI features, is there a mentor, how fast is the team shipping? In a fast-moving field, a slightly lower base on a team that ships and teaches can beat a higher base on a team that parks you in a backlog.
Why it works: it stays warm and collaborative (never adversarial), anchors the ask on external market data plus specific portfolio evidence rather than need, names a single concrete number, and leaves a flexible path to yes. Practice this live — the AI role-plays a recruiter who pushes back ("that's above our band") so you can rehearse holding the number gracefully or trading base for a sign-on.
How it fits together
Lessons teach the concept → the matching capstone deliverable makes you do it on the Support Copilot (and the AI reviews it against a rubric) → interview drills rehearse explaining it → career services package it for the job hunt. That loop — learn, do, get feedback, present — is what makes this a bootcamp rather than a reading list, and it's why all six artifacts are about one feature: you walk into the interview with a single story you can defend end-to-end.