Overview
BeazAtWork is an Ontario job-posting compliance platform built to help job seekers analyze postings before applying and help employers understand and meet new regulatory requirements.
The platform converts legislation into clear, actionable checks using an AI-assisted workflow. It was researched, designed, built, tested, and launched over a four-month delivery cycle.
01 — The Opportunity
Ontario rewrote the rules for job postings overnight
Ontario introduced new job-posting requirements that created a clear need for tools that help employers stay compliant and help job seekers avoid misleading or unsafe postings.
Expected compensation or a salary range must be included.
Any use of artificial intelligence in screening or selecting applicants must be disclosed.
Employers must confirm whether a real vacancy exists.
Employers are prohibited from requiring Canadian experience in any posting or application form.
Positions with compensation over $200,000 do not require salary disclosure.
Employers must provide required information to applicants within 45 days.
Maximum fine for non-compliant employers
Employee threshold that triggers the rules, except Crown corporations
This created an opportunity to build a platform that translates regulatory requirements into clear, actionable guidance, helping companies stay compliant and protecting job seekers from scams.
02 — Role
One builder, seven hats
From regulatory research to shipped product, the same person owned every layer of delivery.
Responsibilities Held
- Product Owner
- Implementation Lead
- Builder
- QA Specialist
- Technical Delivery & Integration
- Workflow & Process Design
- Compliance Research & Translation
- AI-Assisted Development & Review
The Solution
A fully AI-assisted SaaS platform that analyzes job postings, identifies compliance gaps, and provides step-by-step corrections. The system supports both job seekers and employers with clear, structured guidance.
For Job Seekers
- Analyze job postings before applying
- Understand their rights under Ontario’s new rules
- Recognize red flags and potential scam postings
- Know where and how to report fake or misleading jobs: PubliclyAdvertisedJP@ontario.ca
For Employers
- Learn the latest compliance requirements
- Avoid fines of up to $500,000
- Use BeazAtWork’s custom-built template to vet job descriptions before posting
- Ensure postings meet disclosure rules and avoid prohibited requirements
Trust the process. Verify the output.
“Build fast, validate continuously, and keep the compliance logic transparent.”
Delivery Approach
- Mapped regulatory requirements into structured rules.
- Designed a lightweight UX for fast scanning and correction.
- Built AI logic to interpret postings and flag violations.
- Implemented a boutique interface aligned with the brand.
Outcome
The platform is live, functional, and monetization-ready. It provides instant compliance checks, reduces ambiguity, and supports transparent hiring practices.
Project Documentation
GitHub README
View the project documentation.
Key Lessons
I Learned How Real Systems Fit Together
I used FastAPI, PostgreSQL, Clerk, Stripe, WordPress, Railway, GitHub, and multiple APIs. The biggest lesson: building a real product means understanding authentication, payments, databases, hosting, frontend, and backend logic.
I Learned That AI Speeds You Up — If You Control It
I used ChatGPT, Claude, and Copilot throughout planning, coding, testing, and debugging with a zero-trust approach. AI accelerated the work, but only when I verified every assumption. A simple mistake could cost days of debugging.
I Learned That Testing Is a Non-Negotiable Skill
I used pytest, Thunder Client, environment isolation, and structured E2E testing. Issues often appeared several steps later, especially in Stripe payment integration and API chains. An introduction to OWASP changed everything.
I Learned How Multi-Model Peer Review Improves Quality
I used multiple AI models as a virtual peer-review team. One generated code, others critiqued, tested, or rewrote it. Comparing their outputs reduced errors, improved architecture decisions, and kept delivery moving even when one model got stuck.