Paystack to Meta webhook
A PHP webhook that verifies Paystack's HMAC signature, hashes customer email, phone and name, and sends a Purchase event to Meta's Conversions API. It fixed broken ad attribution on a live store.
Trained as a mechanical engineer, I now design and ship working software for real businesses: an AI web app that writes research drafts for paying students, payment-to-ad tracking that Meta verifies, and the ad campaigns that bring in the buyers.
A self-serve product for university students. A student enters a topic, pays through Paystack, and gets a topic-specific project outline or a 40–60 page Chapter 1–3 first draft, with real published studies cited in the literature review. I built the app, the payment and tracking layer, and the Meta ad campaign that sells it.
Five creative angles on Facebook and Instagram, feed and story sizes.
Outline or full draft, a free Chapter One preview, support on email and WhatsApp.
Payment verified on the server, with the amount checked against the tier.
Gemini writes the document in 7 stages, in the background.
Rendered document, Word download, and an email link when it's ready.
One long request would hit the model's output limit and cut chapters off mid-sentence. So each order runs as a queue of smaller jobs, and every stage is saved the moment it finishes.
A cron worker finishes documents in the background. Per-order file locks stop the worker and an open page from writing the same stage twice, so no stage is paid for twice.
Each stage retries three times. If API credit runs out, the order pauses, emails me, and resumes by itself after a top-up.
Every study the model proposes is matched against the pages it actually retrieved. Anything that doesn't match is removed, and the reference list only includes studies cited in the text.
A password-protected admin page shows each order's status, real API cost and search count, with a one-click Resume.
I traced it to web search running inside the writing calls. Every search result was re-read on every later step in the same request, so input tokens compounded with each search. I moved all research into a single step with a fixed search limit, on a cheaper model, and wrote Chapter Two from that checked list.
I set up tracking before spending on ads, so Meta could optimise for real purchases. Browser and server events share the Paystack reference as an ID, so each sale is counted once. Then I launched a sales campaign with five angles and set rules for cutting losers and scaling winners.
| Unit economics | Outline | Chapter 1–3 |
|---|---|---|
| Price | ₦2,500 | ₦13,500 |
| Paystack fee | ₦137.50 | ₦302.50 |
| Left before AI and ad costs | ₦2,362.50 | ₦13,197.50 |
Chapter 1–3 shown at the BSc to MSc price; PhD is ₦15,000. Break-even cost per sale sets the kill and scale rules for each ad.
The payment-to-ad tracking work that the flagship build grew out of. Each one runs on real money.
A PHP webhook that verifies Paystack's HMAC signature, hashes customer email, phone and name, and sends a Purchase event to Meta's Conversions API. It fixed broken ad attribution on a live store.
The same webhook rebuilt as one multi-tenant endpoint. Each client adds one URL in their Paystack settings. I never need access to their server. It's sold as a done-for-you service.
I first sent sales from Selar to Meta with Zapier. Selar only exposed the buyer's email, which kept match quality low, so I moved the store to Paystack and wrote the webhook above.
What goes into the work, and where I've used it.
| Item | Component | Specification |
|---|---|---|
| C-01 | Web development | PHP and JavaScript, REST APIs, webhooks, background job queues with cron, file locking, cPanel hosting |
| C-02 | AI integration | Google Gemini on Vertex AI, Anthropic Claude API, prompt engineering, multi-stage generation, web-search grounding, cost tracking per call |
| C-03 | Payments and tracking | Paystack, Meta Pixel and Conversions API, event deduplication, hashed customer matching, Events Manager |
| C-04 | Growth marketing | Meta Ads strategy, ad creative and copywriting, break-even and cost-per-acquisition modelling, sales funnels |
| C-05 | Automation | Zapier, ManyChat, programmatic PDF and Word document production |
| C-06 | Research and writing | Scoping and systematic reviews (PRISMA), research proposals, academic consulting |
| C-07 | Engineering and analysis | Mechanical troubleshooting, precision measurement and QC, root-cause analysis, data analysis |
| C-08 | Tools | WordPress, Meta Business Suite, Canva, Google Workspace, Microsoft Office |
Research analyst and consulting support for postgraduate clients, including PRISMA systematic reviews and full proposals. Now also sells self-serve outlines and Chapter 1–3 drafts through the AI web app.
Digital wellness guides for Nigerian women and diaspora communities, published under the pen name Adeife Abolaji and the Adeife Speaks blog.
Done-for-you Paystack to Meta Conversions API setup for sellers running Facebook and Instagram ads, fixing broken conversion tracking at the source.
A web studio at gen.provault.online that takes ebook sellers from product idea to PDF guide, sales page and ad pack. Licensed to online community owners, whose members sign up and pay inside it.
| Dates | Role |
|---|---|
| 2025 – now | Research Analyst, Academic Consultant & DeveloperDamiiotye Research Consult |
| 2025 – now | Founder & Digital Operations LeadDamiiotye Wellness Products |
| May – Nov 2023 | Machine Shop InternSumal Foods Limited, Ibadan |
| Jul – Sep 2022 | Auto Shop InternBless Tayuta Automobile Repair Shop, Akure |
| Jul 2021 – Sep 2022 | Auto Shop InternDouble Crown Automobile Company, Ibadan |
| 2019 – 2024 | B.Eng, Mechanical EngineeringLead City University, Ibadan |
Open to developer and integration roles, AI-assisted development work, and freelance tracking or funnel builds.