What should a Hume learner know before trusting an AI model's score?
Where the score came from. A model can look accurate because one of its inputs quietly contains the answer, which is called target leakage. Our Hume project shows it with real data: 49 government schools, a 98% score, and a model that learned nothing about schools.
Hume City Council takes in both Broadmeadows and Sunbury, and the 2025 state file lists 49 government schools inside it. We asked a simple question of that list: can a program tell which schools are secondary? One version gets 48 of 49 right. It is also a model that has learned nothing, and seeing why is the lesson.
Facts last verified 10 October 2026. Teaching is online; no Hume branch is claimed.
Each course below builds towards real models and real data, with honest testing. The free lesson tells us which one fits.

Machine learning from first principles, with train and test splits done properly.
See the syllabus
Python, tables and charts on public data sets.
See the syllabus
Models, evaluation and the mistakes that inflate scores.
See the syllabus
First Python and simple AI ideas for younger learners.
See the syllabusBrowse the course atlas for more than one hundred options and use the coding roadmap to check prerequisites.
The four we are known for
These run underneath everything above. Every one is live and online, placed by ability rather than by age, and the first class is free.

Python, web and AI projects where the learner still owns the thinking.
See the syllabus
Automate the work you already do, then let AI carry part of it.
See the syllabus
Train a model, read what it learned, and be able to say why it is wrong.
See the syllabus
Run AI coding agents on real work without losing control of the codebase.
See the syllabusAll values are for Hume (ABS LGA23270) as printed on the 2021 Census QuickStats page, which we read on 10 October 2026.
| Measure | Hume |
|---|---|
| Persons | 243,901 |
| Median age | 33 |
| Families | 63,644 |
| Private dwellings | 83,928 |
| Household size, average | 3.1 |
| Motor vehicles per dwelling, average | 2 |
| Indicator | Hume | Victoria |
|---|---|---|
| Dwellings that are separate houses | 87.8% (66,487) | 73.4% |
| Dwellings that are flats | 3.0% (2,242) | 12.1% |
| Holding a bachelor degree or higher | 20.2% (38,291) | 29.2% |
| Employed as professionals | 16.8% (17,264) | 25.0% |
There were 18,851 Hume children aged 0 to 4, which is 7.7% of residents.
18,924 residents, 7.8%, were aged 5 to 9.
Census night came during a Melbourne lockdown. 17.8% of employed Hume residents (18,302) said they worked at home that day.
QuickStats rounds percentages and applies small random adjustments to protect privacy, so we print each number as published and never add groups together to make new totals.
The task: label each of Hume's 49 government schools in the 2025 file as secondary or not secondary. Nine are secondary. We tried four rules and scored every one on Hume, then checked the name rules on the other 854 government schools in metropolitan Melbourne councils.
| Rule | Hume (49) | Other metro councils (854) |
|---|---|---|
| Always say "not secondary" | 40 right, 81.6% | 688 right, 80.6% |
| Name contains "College" | 48 right, 98.0% | 785 right, 91.9% |
| Name contains "Secondary College" or "High School" | 47 right, 95.9% | 817 right, 95.7% |
| Five nearest schools by location vote | 40 right, 81.6% | not scored |
A school's name is chosen because of what kind of school it is. Feeding the name to a model hands it the label in disguise. That is target leakage: the score measures the naming convention, not any pattern a model discovered.
The "College" rule scored 98.0% in Hume but 91.9% across the rest of Melbourne. Its one Hume miss is Mount Ridley P-12 College, which teaches primary and secondary years.
Location carries no leak. The five-nearest-schools vote, trained on the other councils, never once predicted "secondary" in Hume and matched the do-nothing baseline at 81.6%. Here the simple baseline is as good as it gets.
Would this input exist before we knew the answer? A school name is decided alongside the school type, so it fails. That single question catches most leakage before a model is ever trained.
The two misses for the stricter name rule are Sunbury College and Roxburgh College, both secondary schools without "Secondary" in their names. Names follow habits, and habits have exceptions. Specific-purpose schools are included in the 49 but not named here.
Learners build these checks into every project, starting with spreadsheets and ending with Python.
| Check | What it catches |
|---|---|
| Compare with a do-nothing baseline | A model that barely beats "always the common answer". |
| Drop one feature at a time | A single input that carries almost the whole score. |
| Score on a different place or time | A rule that only fits the data it was built on. |
The scikit-learn documentation describes the same trap in its guide to common pitfalls: information that would not be available when the model is really used must stay out of training.
Play a guessing game where one clue gives the answer away, then talk about why that clue is unfair.
Rebuild the four rules in Python and find the three schools that break them.
Audit a model at work for any column that is filled in after the outcome is known.
Counts taken from the state school locations list and the department's station entries file.
Hume has 34 primary, 9 secondary, 1 combined primary and secondary, and 5 specific-purpose government schools in the 2025 list.
Sunbury, Broadmeadows, Roxburgh Park, Craigieburn, Coolaroo and Upfield fall inside the boundary; Broadmeadows logged 878,300 entries in FY24-25.
Term 1 at Victorian government schools starts on Thursday 28 January 2027 and ends on Thursday 25 March.
Which Hume
This page follows the ABS outline of Hume City Council, about 503.8 square kilometres. School addresses inside it include Broadmeadows, Craigieburn, Greenvale, Mickleham, Roxburgh Park and Sunbury.
Starting points by age; the trial lesson sets the real one.
Scratch projects that sort, count and decide.
Scratch Coding for KidsAI Literacy for Kids (Ages 8 to 14)Typed code, simple data and how a program learns from examples.
Python and AI for KidsMaths Through CodingReal data sets, models and honest evaluation.
AI and Machine Learning for TeensData Science for TeensMachine learning and generative AI for study and work.
AI and ML MasterclassGenerative AI CourseThey can, and the leak will still be there.
Give an AI assistant the school file and ask for the most accurate classifier. A tool that is optimising for accuracy has every reason to use the name column, and it may well report a score near 98% without a word of warning.
Spotting that takes a person who understands where the data came from. A learner who has seen the "College" rule fall from 98.0% to 91.9% knows to ask what each column means before trusting any number, generated or hand made.
In Hume classes, AI speeds up the code, and the learner stays in charge of the judgement. The longer argument is in why teenagers should still learn to code in 2026.
Everything happens on a video call, whether the learner is in Sunbury, Campbellfield or Greenvale.
The learner shares their screen; the teacher talks through the learner's own code.
Lessons follow the Victorian school term pattern, with breaks in the holidays.
A complete trial lesson with no charge, followed by a course suggestion.
Fifteen to twenty learners grouped by level, from different suburbs and countries.
Two scheduled lessons weekly, so roughly eight in a month.
A 5pm Hume lesson is 11:30am India time while Melbourne is on daylight saving, and 12:30pm after it ends.
Clock changes
Melbourne daylight saving runs from Sunday 4 October 2026 until Sunday 4 April 2027. India stays on one time all year, so we confirm every booking in Melbourne time.
One price list for every family outside India.
The first lesson costs nothing and ends with a recommendation.
Group classes: a month of lessons, about eight, in a group of fifteen to twenty.
One-to-one classes: a month of lessons, about eight, with a teacher to themselves.
Hume families are billed in US dollars. Nothing is paid until the trial has settled the course and a weekly time; rules for holidays and missed lessons are on the pricing page.
Rated 4.9 across 547 Google reviews. These are real reviews, reproduced as written.
★★★★★
"The one step solution for my son. Modern Age Coders make learning coding so simple that kids love it. The teachers explain complex concepts clearly with practical exercises and interactive content."
Ria Mukherjee
Parent
★★★★★
"Modern Age Coders has been a game-changer for me. I struggled to grasp IT concepts and coding before joining, but their classes transformed everything. I can now confidently write complex programs with ease."
Samriddha Mondal
Student
★★★★★
"One of the most wonderful education centres out there. Education is not limited to school syllabus but focuses on skill development."
Vansh Agarwal
Student
★★★★★
"My child Dhairya is really enjoying the Modern Age Coders classes. This is his first online class and he eagerly looks forward to it. I can already see his improvement, and the teachers are very cooperative."
Sonam Oswal
Parent of Dhairya
★★★★★
"Modern Age Coders have wonderful teachers who teach in a clear, easy and practical way. The teacher boosts students' confidence and inspires them to learn without hesitation."
Sonu Goyal
Parent
★★★★★
"I highly recommend this computer coding class! The teachers are incredibly knowledgeable and passionate about coding."
Ritu Kedia
Parent
Share the learner's age and what they like doing. A seven-year-old might build a sorting game in Scratch; a sixteen-year-old might rebuild the school classifier.
Questions Hume parents and adult learners put to us.
Hume City Council had 243,901 residents at the 2021 Census, with a median age of 33.
Target leakage is when a model's input contains information about the answer that would not really be known at prediction time. The model scores well in testing and fails in use.
Compare against a simple baseline, remove one input at a time, and test on different data. A single input that carries the whole score is the usual sign.
The 2025 Victorian school locations file lists 49 government schools in Hume, nine of them secondary.
Ideas about how computers learn from examples suit children from about seven. Building real models in Python usually starts around 13.
Yes. Adults can take machine learning or generative AI courses, in groups or one to one.
Only online. We have no Australian premises, and our listed phone number is an Indian one.
The first lesson is free. Group classes cost USD 100 a month for two live lessons a week, about eight a month, in groups of fifteen to twenty, and one-to-one classes cost USD 150 a month. You agree the course and time before paying.
By level and pace, with fifteen to twenty learners in each, rather than by age or suburb. One-to-one lessons cover any timing a group cannot.
The school counts come from the Victorian Government's 2025 school locations file. The classifier scores are our own calculation.
See the Melbourne page for the wider city, the Victoria guide for the state, or coding for kids in Melbourne for younger learners.