Two Populations, One Platform — Software Delivery at Wineshipping
Wineshipping is the logistics backbone behind some of the best wine in the country — serving major distributors, top vintners, and wine clubs on one end, and a long tail of about a thousand small but serious producers on the other. When I joined as Director of Software Delivery, those two populations were using the same platform in very different ways, and the system was starting to feel it.
Two Populations, One Platform
The whales — large distributors, established clubs, name-label vintners — needed reliability at volume and integration depth: webhooks for order events, real-time status APIs, and direct connections into their own fulfillment systems. The long tail needed something different: low friction and zero technical overhead. An Excel copy-paste import on a page was often the right answer — and building that well is its own discipline.
We leaned into that distinction explicitly rather than building one experience that served neither group well. The platform got integration-ready for clients who wanted it and stayed approachable for everyone who didn't.
Making the Monolith Service-Aware
The core system was a working monolith — and a big rewrite wasn't the answer. Instead, we identified the bounded contexts where scale and change pressure were highest, and gave those domains clear interfaces without disconnecting them from the whole. Think of it less as decomposing a system and more as teaching it where its own boundaries are.
The highest-leverage target was order status. Clients polled it constantly. We extracted, cached, and scaled just that capability — and dropped load on the core system by an order of magnitude. The monolith kept running the operation; the extracted layer absorbed the read traffic.
Inside the Warehouse
The platform work extended beyond the API surface into warehouse operations. We generated optimized picklists for fulfillment staff, routing their picks to minimize travel time through the warehouse. We also tackled box selection: given a variable set of bottle configurations per order, find the combination of box sizes that ships the order without wasted void fill or unnecessary splits. A real bin-packing problem, solved per-order at pack time.
These aren't glamorous features, but they compound. Faster picks, better box selection, and fewer exceptions add up quickly across thousands of shipments a day.
The New Product Hiding In Plain Sight
As the shipper, Wineshipping sits at an unusual intersection: we knew the producer, the distributor, the varietal, the retail price (shipping insurance requires declared value), and the buyer's address — enrichable with public demographic data at the zip-code level. That's a surprisingly complete picture of a wine purchase.
We built data products around that view: recommendation signals based on what similar buyers in similar zip codes had ordered, producer and varietal affinity modeling, and purchase-pattern analysis that could inform both marketing and inventory decisions for our clients. The dataset had one conspicuous gap — we had no idea whether the buyer thought it tasted good. But it turns out you can do quite a lot with everything except that.
Global Teams, Local Accountability
I led delivery across US, Ukraine, and India-based teams. The usual time-zone challenges applied, but the bigger challenge was maintaining quality when a production incident during harvest season is measured in pallets, not pixels. Clear ownership boundaries and well-defined interfaces between teams — the same discipline we applied to the software — kept velocity high without requiring everyone on a call at the same time.
Pragmatic by Design
Wineshipping didn't need architectural fireworks. It needed software that held up through the holiday season — year after year. The best modernization is the kind that improves the system while keeping the operation running, and that's what we built.
