Contact

Work

A few of the teams I’ve joined, what they were up against, and what changed.

Swisscom

Seventy microservices behind a software-defined corporate network

Technical Lead & Senior Software Engineer · Bern, Switzerland · 2023 – 2026

Situation

Swisscom offers businesses a software-defined network (SDN) that overlays their corporate WAN on Swisscom’s network. Because an SDN is built from virtual network functions rather than physical devices, setting up and reconfiguring a company’s network is faster, easier and more affordable. A 15-person team built and operated the roughly 70 microservices that provision it, and the programme’s largest epics needed breaking down before the team could move on them.

What I did

I was hired as Scrum Master and ended up in the code. Along the way I took on DevOps and Product Owner duties wherever they improved the team’s flow, including decomposing and planning several of the programme’s largest epics.

Result

Features kept shipping into a live service that businesses use to run their corporate networks, with the big epics turned into work the team could actually pick up.

Java · Spring · Domain-Driven Design · Python · Pandas · Scikit-learn · Oracle · MongoDB · GitLab · Claude Code · Copilot

Vantage Elevation

One order for a complete elevator

Contract Team Lead · New York · 2018 – 2021

Situation

A private-equity buyer had acquired several elevator parts manufacturers, each with its own configure-price-quote (CPQ) system. A customer who wanted an elevator had to assemble it vendor by vendor.

What I did

I replaced the separate systems with one integrated CPQ platform, and set up AWS-based CI/CD that delivered both into the VPC and onto Vantage’s own premises. I arrived as an engineer and ended up working across the table from the CIO and CFO.

Result

Customers could order a complete elevator in one place instead of piecing it together from several vendors.

Java · Spring · Hibernate · Tacton CPQ · Cucumber · Postgres · AWS · Terraform

Threatray

Making a malware-analysis pipeline easy to change

Contract Senior Engineer · Biel, Switzerland · 2021 – 2022

Situation

A six-engineer startup was applying machine learning to malware DNA analysis: a sophisticated microservices pipeline for dissecting samples at scale. Key components had become hard to change, and many services could only be run inside Docker, so every change meant a rebuild and a wait.

What I did

I reworked the service boundaries toward a hexagonal architecture, and built a local development and debugging setup that let engineers run and debug a single service on its own.

Result

The key components became easy to change, and rebuild-and-wait dropped to seconds.

Python · Kubernetes · Docker · Keycloak · RabbitMQ · Postgres

Commodity futures trading

Trading against our own risk model

Portfolio Risk Decision Support · New York · 2014 – 2016

Situation

A two-person team needed a way to decide how much risk to take in a commodity futures portfolio, and to trust that decision with real money.

What I did

We built a risk decision-support system using Markowitz portfolio theory, principal component analysis and Monte Carlo simulation, and traded against it ourselves.

Result

As confidence in the model grew, we scaled from $1,000 to $100,000 of daily value at risk.

Java · Python · Pandas · Jupyter · Scikit-learn · Bloomberg · R

TalkTalk

Thirteen years of continuous release

Customer Portal & Point-of-Sale · Zug, Switzerland · 2004 – 2017

Situation

The retail channel needed software to run on: a customer portal and point of sale across web, tablets and passport-scanning devices, connected to several suppliers’ systems.

What I did

I built and evolved that software and the back ends integrating it with the suppliers.

Result

The retail channel ran on it through thirteen years of continuous release.

C# · .NET · JavaScript ES6 · Vue.js · WordPress · PHP · SQL Server

NextBus

From device firmware to city-scale architecture for public transit

Software Architect · Berkeley, California · 2003

Situation

NextBus built passenger information systems for public transit: on-vehicle devices that tracked buses in real time and fed arrival information to signs at stops. The vehicle devices needed to be reprogrammable over the air via cellular, and the work called for someone who could write that firmware and also own the architecture at city scale. The goal was to win the Chicago Transit Authority contract — one of the largest transit agencies in the United States.

What I did

I wrote the firmware for the on-vehicle devices: over-the-air reprogramming via cellular, routing to other peripherals on the bus, and a new user experience for the drivers. For Chicago I designed the system architecture, contributed to winning the bid, managed the project, and implemented key parts of the upgrade that brought Chicago’s thousands of buses onto the NextBus platform.

Result

NextBus won the Chicago contract and successfully ran a proof-of-concept covering Chicago’s bus fleet — one of the largest in the United States.

Java · Postgres

Have more work than your team can finish?

Let’s talk