Seven days to a nationwide COVID testing platform

In November 2020, M3 Groupe was standing up COVID-19 screening centers across Geneva to test the population at scale. They needed a booking platform that could take that load from day one. An earlier attempt by an external team had been running for months and had not reached production. I was brought in, restarted the project from scratch, and had the platform live in seven days.

The situation

The clock was the hardest constraint. Testing capacity had to scale immediately, and every day without an online booking system meant saturated phone lines and centers running below capacity. The project had already consumed months with a previous team without going live - so on top of building fast, I had to rebuild trust that it could ship at all.

I took full ownership of the whole system and rebuilt it from a blank page. That week was relentless: long days, very little sleep. Seven days later the platform was in production and taking real bookings.

What I built - the patient side

The public app had to work for the entire population, on a phone, in under a minute:

• Interactive questionnaire to guide each patient through eligibility and the right test path.
• Slot-based reservation system built on pre-generated appointment slots per center, so capacity was explicit and could never be double-booked under load.
• Email notifications confirming each appointment automatically.
• QR code carrying the reservation number, which became the patient’s key at the center.

What I built - the back-office (where the real value was)

The part that made the operation actually run was the administration back-office, built into the same front-end for the medical and center staff:

• QR code scanning to pull up a patient instantly on arrival.
• Patient identification and follow-up through the full testing workflow.
• On-the-spot label printing - a tracking label for each patient, so samples and results never got mixed up at volume.
• LAMAL insurance interface to feed reimbursement, so the centers got paid quickly instead of drowning in paperwork.

This is what separated a booking form from a real mass-testing operation: the platform managed patients end to end, from the slot they picked at home to the reimbursement their insurer processed afterwards.

The stack

I chose tools I could move fast in without giving up reliability:

• Front-end: React (patient app + back-office in one codebase).
• Backend: Node.js with an Express web server.
• Infrastructure: AWS, with an architecture built to stay available under peak national demand.
• Database: PostgreSQL on Amazon Aurora - a managed, highly available data layer that stayed consistent while thousands of people booked in the same window.

The outcome

The platform climbed to several hundred thousand patients within weeks, passed a million after a few months, and kept going. It held up under peak demand, delivered results securely by QR code, and gave center staff a workflow they could run at volume. I stayed on for months to support and evolve it in production before moving on to a new challenge.

A project that had stalled for months was live in a week - and then carried the load of a national testing effort.

What made it work

One engineer, total ownership. With no time for handoffs, I owned the entire stack - front-end, back-office, API, database, and infrastructure. Decisions were made and shipped the same day.

Flexibility where the previous attempt lacked it. Requirements shifted daily as the health situation moved. Rather than fight that, I built to absorb it and shipped the smallest thing that could go live, then hardened it in production.

Managed infrastructure over custom ops. Leaning on AWS and Aurora for high availability let me spend the week on the product and its load behavior instead of reinventing resilience under deadline.

Have a critical project that has to ship - or one that has stalled and needs rescuing? Book a free 30-minute call and let’s talk through it.