AI | BUSINESS | ECOMMERCE
I work at the intersection of technology and business, exploring how AI, digital commerce and intelligent systems can create better ways to operate, sell and grow.
The short version
The short version
I work at the intersection of technology and business, exploring how AI, digital commerce and intelligent systems can create better ways to operate, sell and grow.
Nine years of that, starting with the first Magento role in 2016. Since then: Magento and Adobe Commerce, Shopify, WooCommerce and headless architecture, across B2B, B2C, D2C and marketplace models, on catalogues running to 500K+ SKUs and platforms serving 1M+ monthly users.
The journey
Web, then commerce, then Magento, then Shopify, then headless, then transformation, then AI. Read down the left-hand column and it looks like a technology list. It is not: each move was forced by a problem the previous way of working could not hold.
2010
Web
Six years of formal study in information technology: a diploma from 2010, a degree from 2013. The point at which this stopped being a hobby.
2014
Commerce
The first websites and online stores. Building for other people, and discovering that the interesting problems were never the ones on the page.
2016
Magento
Magento module development and store builds, professionally, from 2016. The platform that turns a catalogue into an architecture problem, which is the lesson that has stayed useful longest.
2018
Shopify & scale
Shopify storefronts for consumer brands, then commerce builds for Saudi retail across grocery, commercial kitchen equipment and fashion. Two very different answers to the same question about where complexity should live.
2020
Headless & performance
Headless frontends over REST and GraphQL, and high-traffic B2B commerce: 90% of order and quote processing automated, page load times cut by 60% through Varnish, Redis, full-page cache and database tuning.
2023
Transformation
Architecture direction across 12+ multi-store Magento 2 platforms carrying 500K+ SKUs and 1M+ monthly users, with five-level approval workflows for development, B2B orders, quotes and vendor management. The work stopped being code and became process.
Now
AI
Where the process work leads. Retrieval, agents and automation aimed at the parts of a commerce operation that absorb people who should be doing something else.
Positions
Technology should create leverage.
If a system does not make a decision better, an experience stronger or an operation faster, it is an expense with a roadmap attached. Novelty is not a business case.
Business strategy comes before technology selection.
Platform debates are almost always architecture debates in disguise, and architecture debates are almost always business-model debates in disguise. Start at the end you can actually answer.
Customer experience is a competitive advantage.
Features get copied within a quarter. The accumulated quality of a journey (how fast, how clear, how little it asks) does not, because it is a hundred small decisions rather than one big one.
AI should solve real problems.
Automating a process nobody has fixed gets you the same bad outcome faster, and at higher cost. Most AI briefs I see are process briefs that have not been recognised yet.
Continuous learning is not optional.
The technology on my CV changed roughly every three years and there is no reason to expect the next three to be gentler. The only durable skill is being able to tell a shift from a fashion.
Method
Five steps, and the order is the entire content. Almost every technology project that fails ran them backwards: it chose the platform first, then went looking for the business case that justified it.
Understand the business
What makes money, what costs money, and which of the two the current system is actually optimised for. This is usually where the real brief turns out to be different from the stated one.
Map the system
The process as it runs, not as it is documented. Where the manual work sits, where the data goes stale, and which step everyone quietly works around.
Choose the technology
Now, and not before. The right platform is the one whose complexity matches the problem: for a small single-store catalogue, that answer is frequently the cheaper one.
Build intelligently
Sequenced so trading continues throughout. In a replatforming the risk is never the build, it is the cutover: URLs, redirects, order history, integrations and the week either side.
Measure the outcome
Against the business number from step one, not against a lab score. A green Lighthouse result on a page that still loses the customer at checkout has measured the wrong thing.
Current focus
Seven subjects, and they are one subject: what happens to a commerce operation when intelligent systems stop being a feature and start being infrastructure.
- AI agents
- LLMs
- AI search
- AI commerce
- Headless commerce
- Business automation
- Digital transformation
Beyond the work
Based in Vadodara, Gujarat. Most of what I know came from building something badly first, which is a slower curriculum than a course and a considerably more durable one.
- Reading
- Primary documentation over summaries. Model cards, platform release notes and the occasional paper, because a second-hand account of a technical change is usually a confident version of a wrong one.
- Building
- Small things, to find out whether an idea survives contact with an API. Most do not, and that is the useful result.
- Writing
- Long-form, on the subjects above. Writing is how I find out whether I actually understand a position or merely agree with it.