Most marketing leaders adopt AI tools. I build the architecture underneath them, and I carry the commercial number that architecture produces. Eleven years, over 50,000 leads and more than 12,000 sales across my own ventures and client work.
Every project here is production infrastructure I designed, wrote and operate. Each figure traces to revenue that was banked or an outcome that was scored. Expand any project for the full architecture.
Figures shown in ₦ are Nigerian naira, where the work was delivered. Approximate rate at time of writing: ₦1,350 to $1.
Problem. A business with real authority on LinkedIn but no owned web presence. Every lead lived inside someone else's platform, with no capture layer and no measurable path from attention to revenue. Growth depended entirely on the founder posting personally, which is a ceiling rather than a strategy.
Built. I recommended moving onto owned infrastructure, then built it: the website, the landing pages, and the funnel connecting paid distribution to conversion. LinkedIn advertising routes into dedicated landing pages, intent is captured there, and qualified prospects return into LinkedIn messaging where the founder's credibility closes them. Paid reach and personal authority made to compound rather than compete.
Problem. A cold start trap. Products without reviews never surface in marketplace search, so they never sell, so they never accumulate reviews. Paid placement was the obvious escape and an expensive one.
Built. I reverse engineered the ranking behaviour and found review count carried disproportionate weight in search placement. Rather than buy visibility, I engineered it: generated demand offline, then deliberately routed those buyers through the marketplace listing instead of selling directly, forgoing immediate margin to capture a ranked transaction. Every buyer was followed up by phone to convert the purchase into a review.
Problem. A live ecommerce operation running paid social with unreliable measurement. Purchase events fired without revenue values attached, so return on ad spend read as zero, conversions were under reported, and no spending decision could be justified with evidence. Budget was being allocated on instinct.
Built. A complete acquisition system covering campaign architecture, creative and copy strategy, and the measurement layer beneath both. Server side conversion tracking implemented through the Conversions API alongside browser pixels, with deduplication so events counted once and revenue values passed correctly. Campaign structure rebuilt so the ad platform could exit its learning phase rather than resetting on every edit.
Problem. Content production, campaign operations, customer support and social engagement each required human attention that did not scale. Output quality varied by who did the work. Hiring was the obvious answer and the wrong one.
Built. A library of fourteen specialised AI agents and skills, each owning a defined function, invoked on demand and running against live production systems. These are not prompt templates. Each carries its own operating rules, tool permissions, guardrails and domain knowledge, and several write directly to production infrastructure.
Problem. In a pay on delivery market, failed deliveries destroy margin. Goods travel, the customer is unreachable or declines, and the cost is absorbed entirely by the business. The operation needed to know which orders were likely to fail before dispatch.
Built. A risk scoring model running a local large language model to assess each order across address findability, customer responsiveness, order characteristics and confirmation signals, returning a risk band and a recommended action. Deployed in shadow mode first, scoring live orders and logging predictions without acting on them, so accuracy could be measured against real outcomes before any automated decision was permitted.
Problem. Revenue reporting stopped at the storefront. There was no reliable view of what an order actually earned once payment fees, landed product cost, delivery and acquisition spend were accounted for. Cash received was being mistaken for profit earned.
Built. A complete commercial system spanning the customer facing store, payment capture, order lifecycle management, inventory, and a reporting layer resolving each order to a true profit figure. Built in Python against live transaction data, with an operations dashboard used daily to run the business.
Paid social strategy, funnel architecture, marketplace and search ranking strategy, conversion rate optimisation, attribution design, cost per acquisition management, creative and copy strategy.
Multi agent system design, agent orchestration, prompt engineering, local and hosted LLM deployment, retrieval augmented generation, content and campaign pipeline automation.
Python, REST API integration, payment gateway integration, server side event tracking, data modelling and reporting, dashboard development, cloud deployment.
Full profit and loss ownership, unit economics, pricing and margin strategy, supplier negotiation and international sourcing, inventory and fulfilment operations.
Across the two largest freelance marketplaces I have generated more than $160,000 in client revenue: over $100,000 on Upwork and over $60,000 on Fiverr. That work spans web development, brand development, store design, WordPress builds and marketing delivery for clients across multiple continents.
My Fiverr profile carries a perfect 5.0 rating across 61 client reviews accumulated over more than five years. Sustaining that across the volume and the timespan is itself the credential: it reflects delivery, not marketing.
Available for growth leadership, AI systems architecture and commercial strategy. Remote, worldwide.