// Selected work

Three systems we
quietly operate.

We don't take on many clients, so the ones we do, we go deep with. Here are three systems we've built and continue to run, in production, today.

01 · Social media operations · SaaS

Growth App

a 985-phone outbound machine

A centralised operations layer running 985 Android devices and 2,760 social media accounts across 4 production servers, built so 16 partners can scale their outreach without ever opening a phone.

985
Android devices
2,760
Accounts managed
4
Production servers
16
Partners onboarded

Engagement

EngagementBuild + ongoing operate, since 2023
Team2 engineers, 1 designer
StackTypeScript, Node, Postgres, ADB, n8n
StatusLive in production

The brief

The team was running social-media growth operations the hard way: humans on individual phones, copy-pasting between spreadsheets, manually rotating accounts when one got rate-limited. It worked, barely, and it didn't scale past about ten devices.

What we built

A central app (elselab.io/growth) that talks to 985 physical Android devices over ADB, distributed across 4 servers. Account pipelines, source accounts, AI-driven search, partner-level segmentation, work-time scheduling, follower / following / story tracking, real-time health checks. Operators see one screen, not a thousand.

The result

16 partners now run their day-to-day operations on top of the system, managing 2,760 social media accounts in parallel. The work-time scheduler runs accounts on their own clocks; reply rates went up because messages get sent at the right moments; the team can onboard a new device or partner in minutes, not days.

02 · E-commerce operations · Switzerland

CatalogOS

a Swiss reseller's full back office in one screen

Operations system for a Swiss reseller running on Europe's largest online marketplace. Order files in, normalised orders + shipments + invoices out. 98,000 products, 9 providers, real-time EUR/CHF pricing, all in one place.

98,453
Products in catalog
78,094
Active SKUs
9
B2B providers
134.6k
CHF / month

Engagement

EngagementBuild + ongoing operate, since 2024
Team2 engineers
StackTypeScript, Postgres, n8n, PDF/Excel pipelines
StatusLive in production

The brief

A Swiss reseller listing on the country's largest online marketplace, running their entire back office out of Excel. Orders arrived as XLSX files, prices had to be re-margined against EUR/CHF every day, nine different providers had different commission structures, and invoicing was a manual Word-to-PDF ritual every week.

What we built

A single operations system that ingests the marketplace's order files, normalises them into a Postgres-backed catalog of 98k+ products, calculates live CHF selling prices off the current FX rate per provider commission, generates shipment manifests and PDF invoices on demand, and keeps a blacklist of 1,055 problem barcodes so they never get re-listed. Plus the side tools (profit calculator, currency converter, bulk product data filler) that the team actually uses every day.

The result

The whole operation runs from one tab. Orders go from Excel inbox to shipped-and-invoiced in minutes, not days. The catalog grew past 98k products without adding people. Every CHF figure on screen recalculates the moment the FX rate shifts. The Monday-morning Excel war is over.

03 · Multi-vendor commerce · Dropship

XML Integration

multi-provider feeds into one storefront, fully automated

An integration layer for a multi-vendor commerce platform. Providers publish XML, sellers map once, cron syncs every minute. Products land on a Shopify-style storefront ready to sell, dropship handoff included. No human in the import loop.

6+
XML providers wired
1×/min
Cron tick frequency
184
Field mappings
0
Manual imports

Engagement

EngagementBuild, 2025
Team1 engineer, 1 designer
StackPHP, Laravel, MySQL, Cron, REST
StatusLive in production

The brief

A multi-vendor commerce platform wanted to let its sellers list bulk inventory from any B2B provider that publishes an XML feed, without the platform team writing a custom integration per provider, and without sellers ever doing a manual import. The catch: every provider's XML schema is different, every seller's commission model is different, and the storefront needed clean, structured product data on the other end.

What we built

A four-piece pipeline. (1) A Providers admin where you paste an XML URL, set a commission %, and you're done. (2) A Mapping UI that connects provider XML fields to the platform's product schema (name, price, stock, variants, attributes), saved per provider. (3) A scheduled sync running off a one-line crontab entry that hits every provider once a minute, normalises, and bulk-pushes products to the seller's storefront. (4) Side tools (Users, Products, API Settings, Clear Cache) wrapped in a clean control panel so the operations team can see what the system is doing without reading logs.

The result

Sellers connect a feed and start selling the same hour. New providers go from "never seen this XML before" to live on the storefront in under fifteen minutes, the time it takes to map the fields once. The cron tick keeps stock and prices fresh without anyone watching. And because the dropship handoff is wired into the same flow, orders forward straight back to the provider with the right margin already taken out.

// 04 — Get in touch

Got something complex?
Let's make it simple.

A short email is enough. Tell us what's hard, what you've tried, and where it hurts. We reply within two working days.