
Aki Wijesundara, PhD
AI Founder | Google AI Accelerator Alum

Manu Jayawardana
Exited AI Founder | Co-Founder, Snapdrum
Most people trying to break into AI Engineer, Forward Deployed Engineer or AI PM roles get stuck at the exact same point: they know they need a project, but they don't know which one. So they build the same RAG chatbot everyone else builds, or a toy demo with no real user, and it doesn't move the needle in an interview.
This resource fixes that. It's ten realistic business problems, pulled directly from the kind of work AI Engineers, FDEs and AI PMs actually get hired to do, each one scoped tight enough to explain in five minutes and real enough that a hiring manager leans in.
For every project you get:
A named persona with a real job and a real company size, not "a business owner"
The workflow that person does by hand today, with a rough time cost
Exactly what the system should do, in plain language
The success metric to build toward, defined before you write a line of code
A difficulty tag, so you can pick something that matches where you are
Three interview questions, one each for an AI Engineer, an FDE and an AI PM interviewer, so you know exactly what a panel will ask about it
The ten projects range from an Easy support ticket triage system to a Hard multi-step research agent, covering support, legal, finance, sales, recruiting, healthcare admin, product, HR, founder ops and education. Pick the one closest to your background, or the one that scares you the most, either works.
You also get a one-page breakdown of the five artefacts every finished AI project needs to actually be interview-ready: a working system, an eval report, a public repo, a demo video, and a case study written for your target role. Most self-taught portfolios stop at the first one. This shows you the gap between "I built something" and "here's proof it works."
And because reading about a project isn't the same as having one, this ships with a working starter kit for the first project, support ticket triage and routing: a 400-row synthetic dataset, a 40-case hand-labelled golden set, and a starter README, so you can be building within the hour instead of staring at a blank repo.
Free
Stop guessing what to build. Ten realistic business problems hiring managers actually want to see.