TAI Lightning Lesson · Free

Job-ready
for AI roles
in 2026

What employers actually look for. And how to prove you have it.

Session

Live · one sitting

You bring

A portfolio you're rebuilding · a laptop

Led by

Aki Wijesundara

Manu Jayawardana

The problem

Knowing how to prompt doesn't make you job-ready.

Signal 1

AI job postings are up.

More roles than ever, across every industry. Every company wants an AI hire. The demand is real and it isn't going away.

Signal 2

Rejection rates are also up.

Employers got smarter about what "AI skills" actually means. The bar rose faster than the pipeline of qualified candidates.

Most candidates are tutorial-rich and project-poor. They finished the courses. They watched the videos. They can define "RAG" and "few-shot" and "MoE." What they cannot show is a system they built that solved a real problem. That gap is where the next thirty live.

By the end of this session

Two pathways. Four signals. A three-project portfolio. And the script for every AI interview.

The receipt: know which pathway you're on (PM or Engineering), know the four signals employers filter for, know how to build a three-project portfolio that gets you hired, and know how to talk about AI work in interviews without sounding like everyone else.

01

2 Pathways. PM or Engineering. Pick one before you build anything.

02

4 Signals. Problem framing. System thinking. Real constraints. Clear documentation.

03

3-Project portfolio. System project. Business project. Edge project. Different signal each.

04

The interview script. STAR for AI. Four sentences that skip the "I'm passionate" trap.

The landscape

Two pathways into AI careers. Pick one before you build anything.

AI Product Managers

AI for non-technical PMs. Evaluate, scope, and ship AI features without writing code. Problem framing, vendor selection, cross-functional leadership.
AI PMs at foundation model companies. Deep technical fluency in model capabilities, evals, and safety. OpenAI, Anthropic, Google DeepMind shaping product direction.
Applied AI PMs. Bridge business and engineering. Own the AI product roadmap. Define use cases, measure ROI, manage model tradeoffs in production.

AI Engineers

AI · ML Engineers. Build and deploy models, training pipelines, and inference systems. System design, MLOps, data pipeline fluency.
Applied AI · Solutions Engineers. Integrate AI into products and workflows. API orchestration, RAG pipelines, multi-agent systems, production constraint management.
ML Platform Engineers. Build the infrastructure layer. Feature stores, model serving, GPU orchestration, CI/CD for ML. Bridges ML and DevOps.
AI Agents · Automation Engineers. Agentic workflows. Tool use, multi-agent orchestration, LangGraph, n8n, autonomous system design.

Pathway first, portfolio second. Building a "generic AI portfolio" that spans both pathways signals confusion, not versatility. Pick the sub-role that lights you up. Build for that.

Overview

Now that you know the landscape. Here's how to get there.

01

The shift. What companies expected in 2024 vs what they expect in 2026. Context: why "can you use ChatGPT" is not the question anymore.

02

The 4 signals employers filter for. The rubric behind every screening call and take-home. Signals: what your resume actually gets read against.

03

How to build projects that get you hired. Tutorial-grade vs job-ready. The three-project portfolio. Portfolio: the artefact that outweighs your degree.

04

How to talk about AI work in interviews. STAR for AI. What never to say. Interviews: the script that separates specifics from vibes.

05

How to prepare for AI hiring loops. The five-stage loop and the prep checklist. Prep: what to do the week before the next application.

Part 01 · The shift

The question changed. Yours needs to change with it.

2024

"Can you use ChatGPT?"

The bar was familiarity. Curiosity was enough. Anyone who could describe an LLM in a sentence had an edge.

2026

"Can you architect a system that uses LLMs reliably at scale?"

The bar is production reliability. Retrieval, guardrails, evals, cost, latency, safety. The word "reliably" is doing the work.

Companies don't want AI enthusiasts. They want AI practitioners who can ship. The demand curve stayed the same. The screening rubric changed. Your resume, projects, and interview script have to match the new rubric.

Part 02 · The 4 signals framework

What employers actually filter for.

01

Problem Framing

Can you explain WHY this needs AI and what success looks like? Not "we used an LLM." Rather "the manual pipeline missed 28% of edge cases; here's why an LLM is the right shape of tool for this."

Signal 1

02

System Thinking

Can you draw the architecture and explain how components connect? Boxes and arrows. Data flow. Where the model sits and what feeds it.

Signal 2

03

Real Constraints

Can you speak to cost, latency, and accuracy tradeoffs? What's the p95 latency budget? What's your cost per resolved query? What's the accuracy floor before you route to a human?

Signal 3

04

Clear Documentation

Can someone understand your project in one quick scan? README that lands the problem, approach, architecture, results, and honest limitations.

Signal 4

Every stage of the hiring loop checks a subset of these four. Resume screen checks 1 and 4. System-design round checks 2 and 3. Take-home checks all four. Nail the four, and every stage moves.

Quick check · halfway pause

We stop. You pick one. We keep going.

What we do

A · Problem Framing. You can explain why.
B · System Thinking. You draw the boxes cleanly.
C · Real Constraints. You know the tradeoffs.
D · Clear Documentation. Your READMEs land.
Drop your letter in chat.

What this is for

Whichever you picked, the other three are the gap.

Most candidates score high on one, medium on one, low on two. The three you didn't pick are your homework for the next week.

The rest of this session gives you the projects and script for all four. Play to your strong signal in the resume. Prove the other three in the portfolio and the interview.

The rule of the session

Companies don't want AI enthusiasts. They want AI practitioners who can ship.

The word "passionate" is a red flag now. The word "shipped" is the tell. Every artefact in your portfolio, every sentence in your interview, every line of your resume needs to demonstrate one thing: you took a real problem, chose an approach, made tradeoffs, and put a working system in front of users. Enthusiasm is table stakes. Delivery is the differentiator.

Part 03 · Portfolio

Tutorial-grade vs. job-ready.

Tutorial-grade

"Built a chatbot with OpenAI API"
No architecture diagram
README says "How to install"
Demo video only
Uses one model, no comparison

Job-ready

"Support agent handling 12 intents, 89% resolution rate"
Clear system diagram with data flow
README: Problem · Approach · Architecture · Results · Limitations
Live deployment or reproducible eval
Compares approaches, explains why one was chosen

The word that changed everything: "12 intents, 89% resolution." Numbers turn a tutorial project into evidence. Every project in your portfolio needs a number attached to it, and the number needs to be measured, not estimated.

Part 03 · The 3-project portfolio

Three projects. Three different signals.

Project 1

The System Project

Shows you can build.

Multi-agent workflow. RAG pipeline with evaluation. Deployed tool with monitoring. The one where the diagram matters.

Signal · System Thinking

Project 2

The Business Project

Shows you can think.

AI audit of a real workflow. Cost-benefit analysis. Implementation plan with numbers. The one where the reasoning matters.

Signal · Problem Framing

Project 3

The Edge Project

Shows you go deep.

Fine-tuning for a niche domain. Eval framework. Voice agents. Something you're passionate about and can defend the depth of.

Signal · Depth · Craft

Not four projects. Three. A portfolio of ten shallow projects reads worse than three well-shaped ones. Pick one from each category. Ship all three. Interlink them by adding the artefacts of one to the README of the next.

Part 03 · Your README is your resume

A hiring manager will spend one quick scan on your project.

01

Problem. What you solved and why it matters. One paragraph. Facts only. Include the number that made this worth solving.

02

Approach. What you tried and why. Include what you rejected. Rejection reasoning is where hiring managers see judgement.

03

Architecture. System diagram showing components and data flow. Boxes and arrows. Where the model sits. What feeds it. What consumes its output.

04

Results. Numbers, not vibes. Accuracy, latency, cost. Before and after. What moved.

05

Limitations. What you would improve. This section shows maturity. Every project you've ever shipped has a limitation. Name yours.

The Limitations section is the differentiator. Tutorial-grade projects don't have one. Every senior engineer wrote one for every project they ever shipped. Adding it signals you know your work has edges. Absence signals you don't.

Part 04 · STAR for AI

The interview script that always works.

S

System Context. "We had a document pipeline handling 10K PDFs a day at 72% accuracy." Set the scene. Include the number. Include the volume.

T

Technical Constraint. "Needed 90% accuracy without exceeding $0.05 per doc." Show that you knew the tradeoffs before you touched a model.

A

Architectural Decision. "Hybrid approach: rule-based for structured fields, LLM for unstructured, confidence threshold routing to human review." The interesting sentence. Name the shape, name what was rejected, name why.

R

Result with Numbers. "91.3% accuracy at $0.04 per doc. Human review load dropped 60%." Numbers again. Both the metric that mattered and one side-effect metric.

Four sentences. One breath. Every AI interview question can be answered in this shape. Practice one full STAR answer for each of your three projects. That is nine sentences of prep, in exchange for every technical round of every AI hiring loop you'll ever run.

Part 04 · Never say this

Three sentences that end interviews.

01

Don't say: "I used LangChain."
Say instead: "I built a retrieval pipeline that chains parsing, embedding, retrieval, and generation with reranking." Naming the library tells them nothing. Naming the shape shows you understand what the library did.

02

Don't say: "I'm passionate about AI."
Say instead: "I built X, measured Y, and learned Z about how these systems work in production." Passion doesn't scale. Specific artefacts do.

03

Don't say: "I completed the Andrew Ng course."
Say instead: "I applied [concept] to solve [problem] and here's what happened." Courses are the price of admission. Application is the differentiator.

The pattern: every "don't say" sentence sounds like something you memorised. Every "say instead" sentence sounds like something you did. Interviewers can hear the difference in the first three words.

Part 05 · The AI hiring loop

Five stages. Each one filters differently.

01

Recruiter Screen. Can you communicate clearly? Do you know what the role is? Filters for: language fit, role clarity, deal-breakers on comp and location.

02

Technical Screen. System design, coding, or case study. How you think, not just what you know. Filters for: reasoning shape. Interviewer wants to see decisions, not answers.

03

Take-Home · Live Build. Code quality, documentation, can you ship? Filters for: signal 4 (documentation) and signal 2 (system thinking). Ship it clean, defend it in the follow-up.

04

Hiring Manager · Cross-Functional. Can you explain tradeoffs to non-technical stakeholders? Filters for: whether you can carry the AI conversation into meetings that decide budgets.

05

Culture · Values. Do you learn? Collaborate? Stay curious? Filters for: the human question. Answered by everything you've done in the previous four rounds.

Every stage is a different signal. Prepare for each stage separately. The recruiter screen and the culture round both feel low-stakes. Both are where the most candidates fall out because they're not preparing for signal 1 or the human question.

Part 05 · The prep checklist

Do these before your next application.

01

3 portfolio projects using the rubric (System, Business, Edge). Each with a README that lands the five sections in one quick scan.

02

A "tell me about yourself" answer using STAR for AI. Around two paragraphs. Practiced out loud, timed, refined. Not just written down.

03

Diagram 3 AI systems you've built or studied. On a whiteboard, from memory. Boxes and arrows. Data flow. Where the model sits.

04

Prepare 3 questions showing you understand the company's AI challenges. Read their engineering blog. Read their latest launch. Ask about the tradeoffs they mentioned.

05

Practice explaining one project to a non-technical person under a short elevator ride. The parent-test. If your parent gets it, the hiring manager will too.

This is a one-weekend build. Not a six-month plan. The candidates who do these five prep tasks over one weekend are the ones who convert screens into offers. Everyone else applies and hopes.

What you do this week

Rebuild one existing project. Job-ready shape.

01

Pick one project you've already built. The one closest to the pathway you chose. Don't start from zero. Rebuild the artefact around the four signals.

02

Rewrite the README as Problem · Approach · Architecture · Results · Limitations. Every section under three paragraphs. Every claim has a number attached.

03

Draw the architecture diagram. Boxes, arrows, data flow. Include where the model sits. Excalidraw, Figma, whiteboard photo. Any tool works. Absence loses the interview.

04

Write your STAR-for-AI answer for this project. Four sentences. Say it out loud. Time it against a heartbeat. Too short, you're missing the numbers. Too long, you're rambling.

The receipt: a public repo with the four artefacts (README, diagram, results, limitations) and one paragraph titled "which signal I strengthened and why." Send us the link. We open Week 1 of the Agentic AI Engineering Bootcamp by reading three of yours together.

The habit to take with you

Attach a number to every AI claim you make about yourself.

"Built an AI system" is vibes. "Built an AI system that resolved 89% of intents at $0.04 per query" is evidence. This habit compounds quietly. The candidates whose interview loops convert are the candidates whose every sentence has a number attached. Start with your resume. Then your READMEs. Then your interview script. The habit compounds.

Nine weeks. Build the system project and the edge project inside the cohort. Ship a portfolio you can defend in the interview loop.

Where this leads

Now

Questions in chat

Post the sub-role you're targeting and the one signal you're weakest on. Aki or Manu will read a few aloud and answer live.

Open floor

Later

The rebuild

Send us the rebuilt README, the diagram, and the STAR answer. We reply with a short audio review before the bootcamp starts.

The receipt

Job-readyfor AI rolesin 2026