I build complete web products —and the AI inside them.

I built the Salesforce and HubSpot integrations with real-time data synchronisation, and a 51-agent automation layer sitting on top that creates opportunities, matches partners and generates go-to-market strategy without a human in the loop.

The measurable result was a 95% reduction in errors and a 20–30% productivity gain for the partner teams using it. Those numbers came less from the AI itself than from removing the manual re-keying between systems — the agents are only useful because the sync underneath them is trustworthy.

An eight-month engagement covering rank research, site auditing, SEO issue and opportunity detection, site-structure planning and SEO-friendly content generation — with backlink analysis, an AI watcher and page-speed modules alongside.

The centrepiece is a five-agent automation flow that chains those capabilities into one hands-free workflow. The client's feedback described consistent professionalism, reliability and straightforward communication across the whole collaboration — which on an eight-month build matters more than any single feature.

I built a system that takes one or more source images and produces sixteen differently-proportioned ad formats, using AI background extension to fill the space a crop can't cover — so a wide hero image becomes a tall skyscraper without the subject being cut in half.

Alongside it I built an image editing studio on Fabric.js where users refine the generated output: adjusting layout, adding and styling text, and making per-format modifications before export. The client extended the engagement into a second contract, and both were rated five stars — the second one describing results that exceeded expectations.

I built the product full stack: the website and dashboard interface, the backend, and the integration layer that applies decoration over Dropbox and Google Drive folders without disturbing the underlying files. It includes an AI-powered sticker generator and a marketplace for sticker packs.

This one is a good example of what I mean by honest input rather than order-taking — the client's feedback specifically noted that I contributed ideas that upgraded the app beyond the original brief. On a consumer product where the whole value is delight, that kind of back-and-forth is the job.

I built an MVP that extracts functional areas directly from uploaded Excel files, generates the events and their operational hierarchy from that structure, and lets different roles assign the generated tasks to people. The committee's existing planning document becomes the input rather than something to be re-entered.

I also integrated an AI assistant that suggests tasks and recommends updates to existing ones, and built the client-facing site alongside the platform. Prototypes like this are where honest scoping earns its keep — the goal is proving the concept, not building every feature someone can imagine.