North Bay Scholarships
A Free Scholarship Finder for North Bay Families, Kept Current by AI That Has to Show Its Work
Sector
Education, Community Resource
Project Type
Web App, Data System, AI-Assisted Updates
Timeline
Built and launched October 2026
Tech Stack
Astro, Preact, Tailwind CSS, Vercel (Hosting, Web Analytics), Claude API, Vitest
Website
northbayscholarships.ca
The "Before": Scattered and Hard to Sort
This one started at home. I'm a dad with two teenagers in North Bay, and when we started looking for scholarships and bursaries, I couldn't get a straight answer online. The information was spread across school, foundation and government sites, and it was hard to tell which awards actually applied to us.
The big national scholarship sites don't help much. They're built for the whole country, so small local awards are easy to miss, and every family is left to read the fine print themselves.
The Main Pain Point: Families Don't Need Another List. They Need to Know What Applies to Them.
A parent with a Grade 10 student doesn't have time to read 100 eligibility pages. They need:
- Every award in one place, with local North Bay and Nipissing awards first
- A quick way to see which awards fit their student
- What to do now, and what's coming up month by month
- Information they can trust is current
- No accounts and nothing about their kids stored on someone else's server
What I Built for Families
The site lists 128 awards, 46 of them from North Bay, the Nipissing area and Northern Ontario, followed by Ontario and national awards open to Ontario students. Each award has its own page. The work is in helping a family get from that list to the handful worth their time.
From a List to a Plan
A Three-Minute Interview
A few questions about grade, marks, plans and activities. The matches come back ranked into strong fits, possible fits and awards worth knowing about. One firm rule: a question the family skips never rules an award out.
A Timeline That Says What to Do Now
A month-by-month plan with each award's deadline plus the big dates every family faces: university and college applications, OSAP and RESP. Dates that aren't posted yet are shown as "Expected around," never as confirmed.
A Shortlist That Goes in the Calendar
Families star the awards they want to apply for, then download one calendar file with every deadline. It works with Google, Apple and Outlook calendars.
Plain Language Throughout
Words like bursary and renewable get a short explanation right where they appear, and a glossary covers the rest. No jargon, no fine print to decode.
Private by Design: Answers Never Leave the Device
Families are telling the site about their kids: grades, marks, background, plans. That information has no business on a server, so it never goes to one.
How It Works
Matching Runs in the Browser
Answers are saved in the family's own browser, and the matching happens right there. There are no accounts, and nothing about the student is ever put in a web address.
Families Stay in Control
One button clears everything from the device. To switch from a phone to a laptop, families download their answers as a file and load it on the other device.
Analytics Without Cookies
Visits are counted with Vercel Web Analytics: pages viewed, country and device type. No cookies, nothing that identifies a visitor, and answers are never part of it.
The "Engine": AI That Has to Show Its Work
A scholarship list is only useful if it's current. Amounts change, deadlines move, awards get discontinued. Checking 128 award pages by hand every month isn't realistic, and letting AI rewrite them on its own isn't safe. AI models are good at reading pages and bad at admitting when they're guessing.
So the update splits the job. Claude does the reading. Code checks every claim. A person makes the final call.
The Monthly Update
Every Award Rechecked Each Month
The update reopens every award's official page, reads a watch list of school, foundation and government pages for anything new, and searches for new local awards. Pages that haven't changed since last month are skipped, which keeps the cost down.
No Change Without Proof
Claude proposes changes, but each one must come with a quote from the official page. The script only applies a change when it finds that quote in the page it downloaded itself, and the quote states the new dollar amount or the full date. No quote, no change.
Risky Changes Go to a Person
Eligibility changes, removed amounts and discontinued awards are never applied automatically. They're flagged for review, along with any new award that looks too much like one already listed.
Nothing Goes Live on Its Own
Every update arrives as a pull request with a plain-English summary for review. A public report of each month's changes appears on the site's Updates page, so families can see what changed and when.
A Practical Detail: It Runs on a Mac in the Office
Several of the most important local sources, including Canadore College and Nipissing University, block requests from cloud servers. So the update runs on a schedule on my own Mac instead of in the cloud, in a clean copy of the project, and stops itself if last month's changes haven't been reviewed yet.
The takeaway for business: this is the pattern I'd use for any company that wants AI to keep its data current. Let the model do the reading, have code check every claim against the source, and keep a person on the final call.
Data Rules: Never Guess
- Every award comes from the provider's own website, never from another scholarship list
- No invented amounts or deadlines: anything not confirmed stays blank and is marked as unconfirmed
- Last year's deadlines are shown as last year's, not passed off as this year's
- 1,182 automated tests, including checks on every single award record
- A link checker that tests every award link and official source each month
Performance: Perfect Scores on Mobile
Most families will open this on a phone, often from a link a friend or guidance counsellor shared. The site is built with Astro, so every page ships as plain HTML, and the interactive parts (the interview, results, timeline and shortlist) only load their code on the pages that use them.
Lighthouse (Mobile), Home Page
100
Performance
0.9 s
First Contentful Paint
0.9 s
Largest Contentful Paint
0 ms
Total Blocking Time
1.9 s
Speed Index
0
Cumulative Layout Shift
Accessibility, Best Practices and SEO all score 100.
The Results
128 Awards in One Place
Local awards first, each checked against its official page, with a ranked list and a month-by-month plan for every family.
Perfect Lighthouse Scores
100 across Performance, Accessibility, Best Practices and SEO on mobile.
Nothing About Families Leaves Their Device
No accounts, no cookies, and matching that runs entirely in the browser.
Data That Stays Current Without Guesswork
A monthly AI update that can only change what it can quote, with a person reviewing every change.
The site launched in October 2026, so it's too early for usage numbers. I'll add them here once there's a full school year of data.
Why This Approach Works: Key Differentiators
1. Built from a Real Problem
Every feature came from what my own family needed: what applies to us, what to do now, and what's coming next.
2. Private by Architecture
Instead of promising to protect families' data, the site never collects it.
3. AI with Guardrails
AI does the reading at scale. Code checks every claim against the source page, and a person approves what goes live.
4. Honest About What Isn't Known
Unconfirmed amounts stay blank, expected dates are labelled as expected, and every award links to its official page.
5. Fast and Low-Maintenance
Static pages, no database and no admin login to attack. The only moving part is the monthly update, and it asks before it changes anything.
Final Takeaway
North Bay Scholarships turns a scattered mess of award pages into a short, ranked list and a plan, without asking families for an account or storing anything about their kids. Behind it, a monthly AI update keeps 128 awards current, and it can only change what it can prove from the official source.
// Next Step
Could Your Customers Use
a Tool Like This?
If your customers dig through lists, PDFs or spreadsheets to find what applies to them, a tool like this can do the sorting for them, and AI with guardrails can keep it current.
// More Case Studies
- Morland Real Estate AppraisalsWebsite build and Astro rebuild for a regional appraisal firm.
- Northern Gateway DevelopmentsStrategic communications platform for a major development project.
- Complete Landscaping & PoolsMulti-brand digital growth system across two sister companies.
- Ashley MeghanBrand-first Squarespace redesign built from new brand guidelines.