Session 5 :: Put AI to Work

Today you take on a real professional challenge from your own harbor and produce the kind of work people get hired to do. Everything you need to start is here, that is the floor. How far you take it is up to you, and there is no ceiling.

Start here. Open the Port of LA employment page and pick one tech role that interests you, terminal automation technician, marine electronics, logistics or data analyst. Open ChatGPT.

The real work. Build a complete, honest pathway to that job that a Wilmington student could actually follow: what it pays, what certification or training it needs, which local schools teach it (LA Harbor College, LA Trade-Tech, Harbor Occupational Center), how long it takes, and which employers hire for it. Use AI to draft it, then verify every claim against a real page. Deliver a clean one-page pathway.

Push it further. Compare two roles side by side. Find a real open job posting and list exactly what it demands. Pin down the single cheapest, fastest route in. Draft an outreach message to an employer, or three sharp questions for an informational interview.

Start here. Open the real data: Port of LA monthly container counts and, for the air question, airnow.gov (zip 90744). Paste a slice into ChatGPT.

The real work. State your hypothesis first, then test it against the numbers. Find the real pattern, make a chart, and write a short brief the way an analyst writes for a boss: the finding, the evidence, and one clear recommendation. If you take the air-and-traffic question, pull both datasets and look for a connection.

Push it further. Quantify the trend, how much and how fast. Name what else could explain it. Make a simple forecast for next quarter. Then rewrite your brief for a specific reader: a terminal operations manager, or a public health officer for Wilmington.

Start here. In ChatGPT, build the detector: "You are a scam detector. When I paste a message, list the warning signs and give a verdict: Safe, Careful, or Scam. Explain in Spanish if I ask."

The real work. A tool you cannot measure is a tool you cannot trust. Build a test set: gather or write five real scams and five real safe messages (a normal text from a friend, a real delivery notice). Run all ten. Count how many it gets right and where it fails. Improve it, then test again. The hard part: does it ever call a safe message a scam? That false alarm is the whole job.

Push it further. Measure its false-alarm rate against its missed-scam rate. Try to fool it on purpose. Check it for bias: does it treat Spanish messages or certain senders unfairly? Then write an honest half-page on what it is good at, what it is not, and whether you would ship it.

What you built today, a pathway, an analysis, a tested tool, is real work and a real portfolio piece. Keep it. Next time you take it all the way: build a working app for the harbor and put it live on the internet.