Writing code stopped being the differentiator. What makes the difference is understanding the business behind it.

The product monitors price offers — the brand's and its competitors' — across many marketplaces, 24/7. The hard rule came from the business side: the whole universe of offers has to be collected inside a one-hour window that starts at :00 and closes at :59.

A serial crawler could never fit that window, so the collection scales horizontally on Kubernetes — 100 GB+ machines, 150+ pods. Each collection stores a screenshot as evidence, which is what lets the client audit any offer, and an anti-detection layer keeps the crawlers from being blocked as bots.

It costs more infrastructure and permanent maintenance of the anti-bot evasion. That is a cost I chose knowingly, because the window and the proof are what the client is actually buying.

70+ routes across Brazil, walked by people who visit stores to audit them. The routes had to be optimised without ignoring what field work really looks like.

The problem is a VRPTW, so I modelled it with Google OR-Tools (this was before LLMs), adding penalties, visit priority by coordinate and real capacity limits: working hours, lunch, first and last store, the trip home. Public transport cost came from Google APIs per city, and where there was no data, from the average fare by postal code — so a route could be checked for economic sense, not only geographic sense.

A classic solver means modelling every constraint by hand, which is slower to build. I took that on because the flexibility to tune each parameter mattered more than getting there fast.

Training apps are fragmented by sport. Apex is my attempt to bring multi-sport training and AI coaching into one place. I train — running, strength, trail, conditioning — so I am the user, which removes most of the ramp-up in understanding the domain.

Marketing is treated as part of the product: automated blog content with real sources, social output and an SEO structure built to rank. On the infrastructure side I leaned on managed services — Fly.io and Vercel for hosting, AWS workers, Trigger.dev for background jobs, Supabase for auth, Stripe for payments — so one person could carry the whole thing.

The scope is large for one person, and leaning on automation means also maintaining those pipelines. In exchange I get to learn each of those services deeply, which comes back into the work I do for others.