Topic 10 — Landing an AI-adjacent role

Goal: Use your skills and AI literacy to land an entry-level AI-adjacent role.

Lesson 10.1 — Why the timing is on your side

Bina Velasquez spent eight years teaching high-school biology before she ever wrote a prompt for money. When she started applying for AI work, she expected a wall of people with ten years of experience she could never match. She kept looking for them in the job descriptions. They weren't there.

They couldn't be. The roles she was applying for — AI quality analyst, prompt reviewer, evaluation specialist — barely existed three years ago. Nobody has a decade of "AI-adjacent experience," because the field hasn't been around for a decade. This is the part worth sitting with: a brand-new, fast-growing field has no entrenched traditional path, which means the playing field is unusually level. You aren't behind a crowd of veterans. The crowd is forming right now, and you're in it.

And the demand is real, not hype. In 2026, AI skills show up across a large share of US tech postings, and the count of jobs explicitly requiring AI fluency has grown sharply year over year — more than doubling in many measures. It's the fastest-growing skill category in the US job market.

A field this young doesn't reward the longest résumé. It rewards the person who can show what they do with AI today.

When Trellix Health hired Bina, the founder said the deciding factor wasn't her years — it was that she'd already done the work on her own, before anyone paid her to. That's the door the timing holds open.

Lesson 10.2 — The one sentence that makes you valuable

Here's a trap Bina almost fell into: she tried to compete as an AI person. She studied transformer diagrams, memorized model names, and felt like a fraud next to Pri Reddy, the ML engineer who actually builds the pipelines.

That was the wrong race. Pri can connect a model to a RAG pipeline in an afternoon. What Pri can't do is read a Scribe-drafted clinical note and feel something is off — that the dosage phrasing is grammatical but would make a real physician wince. Bina can. Eight years of grading lab reports trained her to spot a confident, well-formatted, wrong answer at a glance.

That's your edge, and it compresses into one sentence:

Domain knowledge + strong human skills + genuine hands-on AI literacy. "I know this domain and I understand AI."

Break it into its three parts, because you'll lead with this everywhere:

• Domain knowledge. Your prior field is the asset, not the thing to hide. Healthcare, finance, education, law, customer service — every one of them needs people who can judge AI output inside that world. Bina knows what a clinical note should say. Most prompt-writers don't.
• Strong human skills. Clear thinking and writing (the heart of good prompting), patient judgment, attention to detail. These are exactly the teacher's instincts that made Bina good in a classroom.
• Hands-on AI literacy. Not theory — fluency. You can prompt well, spot a hallucination, and evaluate an output against criteria (Topics 4–7).

A career-changer who can say all three honestly is rare. A new grad who can only say the third one is everywhere. Lead with the combination.

Lesson 10.3 — Build proof, not claims

A résumé claims you can do the work. A portfolio shows it. In a field this practical, demonstrated skill beats a stated one every time — and the best part is you can build the proof yourself, for free, this month.

Bina built three small projects before she had a single interview. You want two or three, not ten. Aim for these:

Build something with AI. A custom assistant or a tested set of prompts that solves a real problem in your domain. No-code counts fully — the skill is in the design and testing, not the plumbing. Bina built a small assistant that turned messy biology lecture notes into clean study guides, then documented why each prompt was shaped the way it was.

Do an evaluation project — most people skip it, and it's the most valuable. Take an AI's output on some task, score it systematically against written criteria (accuracy, completeness, tone, safety), and write up what's wrong and how you'd fix it. This is AI Operations work. Bina took thirty AI-generated patient-message replies, scored each one, and wrote a two-page memo: which failed, why, and what prompt change would catch the failure. That memo got her the Trellix interview. Marc Devlin, the AI PM, said it read like something a year-two employee would produce.

Strengthen it with public artifacts. A few thoughtful posts — analyzing a tool, sharing a prompt technique, walking through your evaluation method — make your literacy visible and searchable. And keep the real-work examples: "I used AI to cut my report-writing time in half" is a story a hiring manager remembers.

Shared simply — a personal page, a doc, a short write-up each — two or three of these turn "I'm interested in AI" into "here's what I've already done."

Lesson 10.4 — The résumé that gets the call

Bina's first résumé led with "Biology Teacher, 2018–2026." It got no replies. The problem wasn't her background — it was the order.

Rewrite the top of the page to lead with AI literacy and the relevant skills, then anchor it with domain expertise:

• Name the skills plainly: prompt engineering, AI evaluation and quality, and the AI tools you've actually used — ChatGPT, Claude, a labeling or eval tool, a no-code assistant builder. Name only what's true; an interviewer will ask.
• Pair them with your domain and transferable skills, framed as the one sentence from Lesson 10.2.
• Link the portfolio. The projects do the convincing the bullet points can't.

Then the move that changes everything: quantify. Vague turns to credible the moment a number appears.

Before (vague)	After (quantified)
"Used AI to help with reports"	"Used AI to cut report-writing time ~50%"
"Evaluated AI outputs"	"Scored 30 AI drafts against a 5-point rubric; flagged 8 unsafe ones"
"Good at writing prompts"	"Built a tested prompt set that raised first-draft acceptance to 80%"

You won't have a number for everything. Put one wherever you honestly can. "Reduced X by Y%" is the language hiring managers read fastest.

One reality check so your expectations are calibrated. Entry annotation and trainer contract work runs roughly $15–30/hr; specialized trainers (a real domain credential behind them) command $40–80/hr or about $80–120K; and AI PM or embedded prompt roles commonly land $90–175K in the US in 2026. Many people enter through the accessible evaluation and operations work and grow upward — which is exactly the path Bina is on.

Lesson 10.5 — The interview, and running the hunt like an experiment

Bina's first AI interview wasn't a conversation about her past. Theo Brandt-Okonkwo slid a laptop across the table and said, "Here's a Scribe draft. Walk me through whether you'd trust it." That's the shape of these interviews — practical, not theoretical. Expect three flavors:

• Practical AI tasks. Write or critique a prompt, evaluate an AI response, talk through how you'd handle an AI error. This is Topics 4–7, live. Hands-on fluency is visible in seconds; so is the lack of it. Practice out loud before you go in.
• Judgment and ethics questions. "How would you tell if this output is trustworthy?" Name the real concerns — hallucination, verification against a source, bias, responsible use — and how you'd check. When Bina answered Theo's question, she didn't say "looks fine." She said, "The dosage is unsourced, so I'd flag it for Dr. Marchetti before it reaches a patient." That sentence got her the role.
• Domain + AI scenarios. "How might AI help in your field, and what would you watch for?" Here your domain knowledge stops being a side note and becomes the answer.

Then treat the whole job hunt the way the field itself works: as an experiment. Try an approach, watch what lands, iterate. Bina sent fifteen applications with the old teacher-first résumé and heard nothing. She rewrote the top, led with the portfolio, and the replies started. She didn't get unlucky fifteen times and lucky once — she ran a test, read the result, and changed one variable.

That iterative, evidence-driven instinct is the same one the work rewards. Motivated learners who can show what they do with AI are exactly who companies are hiring right now. That's not encouragement. That's the market.

Worked example — Bina's path onto the Scribe team

Rewind nine months. Bina is still teaching, curious about AI, no industry contacts.

She names her edge. One sentence on an index card by her laptop: biology + clear writing + careful judgment + hands-on AI. Everything she builds points back to it.

She builds three proofs. A no-code assistant that turns lecture notes into study guides. A set of tested grading-feedback prompts. And the project that mattered most — she pulls thirty AI-drafted patient-style message replies, scores each against a five-point rubric (accurate, complete, safe, clear, appropriate tone), and writes a two-page memo on the eight that failed and the prompt fix for each.

She rewrites the résumé. AI literacy and tools up top, biology and the eval memo right under it, every line quantified where honest: "scored 30 drafts, flagged 8 unsafe." Portfolio linked.

She applies like an experiment. First batch, teacher-first framing: silence. She flips the order, leads with the memo, and Trellix Health replies within a week.

She interviews. Marc says the eval memo read like a year-two hire's. Theo hands her a live Scribe draft; she spots the unsourced dosage and says she'd flag it for Dr. Marchetti rather than wave it through. Pri realizes Bina can judge clinical text she herself can't — the exact gap the team needed filled.

She lands it as an AI quality analyst, the accessible on-ramp, and starts growing across the AI-adjacent role family from there. The thing that opened every door wasn't a credential. It was proof she'd already done the work.