Google Professional Machine Learning Engineer Readiness

A coach that teaches by asking — not by lecturing

The Lighthouse Coach is an AI Socratic coach built on Cognosco Learning's Google Professional Machine Learning Engineer Readiness course, in the AI at Work category. It does not play videos at you. It holds a conversation: it asks you a question, listens to your answer, names the misconception it hears, affirms what you got right, and asks the next question — the way a great coach works in the room.

You can fit a model. That is not what this credential tests. In a notebook, with a tidy extract and no deadline, fitting a model is the easiest part of machine learning on Google Cloud. The examination — like the job it certifies — spends most of its weight everywhere else: on whether you framed the right problem before you touched a model; on whether your data can be trusted and your evaluation believed; on whether the features that fed your fitting runs are the same features your live system will see; on whether your serving choice fits the request pattern you actually have; and on whether you will notice, weeks after launch, that the world has moved and your model has quietly gone stale.

It runs on one stipulated case. Devika Rao — Dev to her team — is a machine learning engineer at Sirocco Fresh Logistics, which moves chilled food from six regional depots to 340 grocery stores with 210 refrigerated trucks. Her project is a spoilage-risk model: in the last full quarter, 4.2 percent of chilled shipments arrived with temperature excursions serious enough for the receiving store to reject some or all of the load, each rejected load costs 1,850 dollars on average, and depot managers assign the limited second-check slots — enough for roughly one shipment in five — by a printed rule sheet, based on product type alone, that catches about half the loads that go on to spoil.

Module 1 frames, feeds, and fits honestly. Frame a business problem as a learning problem with a baseline and a success measure stated before any model exists; choose between low-code, managed, and custom approaches by constraint rather than by fashion; treat data as a contract — sourced, checked, and protected against leakage; label, split, and evaluate so the score you report is a score you can defend; engineer features that survive the journey into production; fit without fooling yourself about imbalance, tuning, and the gap; keep an experiment record another engineer could reproduce; and earn the right to ship through gates passed on evidence, not on enthusiasm.

Module 2 ships it and runs it. Serve for the request pattern you actually have — live answers for one record at a time, or bulk answers for a warehouse full of them; scale serving without scaling cost blindly; automate the pipeline end to end so the path from new data to new model is a system and not a favour; orchestrate refitting on evidence rather than on a calendar alone; monitor drift, skew, and slow decay after launch; explain, audit, and govern models other people's decisions depend on; secure the estate; control cost and reliability together; and walk into the examination room ready, knowing your honest standing from scored practice.

It works in two movements. First, Module 1 — Frame It, Feed It, Fit It Honestly — work through Concepts 1–9, finishing with The Design Dossier; then, Module 2 — Ship It, Run It, Prove You Are Ready — work through Concepts 10–18, finishing with The Launch and Operations Plan.

Both capstones produce real artifacts — a complete design dossier for the spoilage-risk model: the frame operations could sign, the approach chosen by constraint, the data contract, the evaluation and the features, the fitting protocol and its record, and the four-panel ship gate with its launch decision form; and a launch and operations plan for Sirocco's second model, the delivery-delay predictor for the customer service desk: its hybrid serving design, its automated pipeline, its refitting and monitoring system, and its governance, security, and cost position — the plan you then rebuild, in the variant, for a model of your own.

The honesty standard is part of the course. This is a readiness course, not a promise: it prepares judgment for the examination room, it guarantees no pass, and it reproduces no actual examination items — every scenario in it is an original teaching scenario. It assumes you have fitted models before and worked on Google Cloud before. What it adds is the level the examination actually samples: the judgment calls, in sequence, under constraint, with the trade-off named out loud.

The course is self-paced and always on — roughly 30 minutes per concept, about 9 hours in full, taken in as many sittings as you like. You can type your answers, or talk out loud: the Voice Coach speaks its questions and listens to your spoken replies. Learners who already judge an area well can test out concept by concept — the bar does not drop; the route just gets shorter.

Google Professional Machine Learning Engineer Readiness is part of The Lighthouse Signature Series — the online counterpart to the one-day, in-person learning Cognosco Learning delivers.