Daniel Coderre, Developer in Bern, Switzerland
Daniel is available for hire
Hire Daniel

Daniel Coderre

Verified Expert  in Engineering

Data Engineer and Software Developer

Bern, Switzerland

Toptal member since October 17, 2022

Bio

Daniel is an experienced data professional with a decade of experience in building analytical systems. He is happiest working on the full stack, from engineering storage systems and processing pipelines to data analysis, presentation, and building front-end dashboards. For the past several years, Daniel has been heavily into machine learning (ML) and has applied it to create value from real-world data. He has a PhD in experimental physics, so ML is not the only tool in his toolbox.

Portfolio

PostFinance
Kubernetes, PySpark, Python, SQL, MongoDB, AngularJS, Machine Learning...
University of Freiburg
Python, C++, Statistics, Physics, Big Data, Electronics, Team Leadership...
University of Bern
Python, C++, Django, Linux Server Administration, MongoDB, Electronics...

Experience

  • Data Processing Automation - 10 years
  • Data Analysis - 10 years
  • Machine Learning - 7 years
  • Python - 7 years
  • C++ - 7 years
  • Remote Team Leadership - 5 years
  • Kubernetes - 3 years
  • PySpark - 3 years

Availability

Part-time

Preferred Environment

Linux, Emacs, Visual Studio Code (VS Code), Jupyter

The most amazing...

...experience I've had was building the DAQ for a dark matter detector, leading an analysis team, and guiding our data from analog cables to final plots.

Work Experience

Senior Data Scientist

2020 - PRESENT
PostFinance
  • Created an ML modeling platform from scratch in PySpark and Kubernetes to automatically perform a complex series of ETLs, build model training sets, normalize inputs, train models, and serve results. It is easily monitored by a web front end.
  • Built several ML models for customer behavior predictions for advertising. These models have performance improvements over human-based methods and bring enormous cost savings. Key technologies used: Spark, Kubernetes, XGBoost, and scikit-learn.
  • Built an ML system for anomaly detection in transaction data using a sparse training set. It has superior performance to software from multiple industry-leading vendors. Key technologies used: Apache Airflow, Kubernetes, Spark, TensorFlow, and XGBoost.
  • Built an ML-on demand live web API serving preprocessed results as well as live time-series predictions. Key technologies: Kubernetes, Flask, and Prophet.
Technologies: Kubernetes, PySpark, Python, SQL, MongoDB, AngularJS, Machine Learning, Neural Networks, APIs, CI/CD Pipelines, GitLab CI/CD, Helm, Data Analysis, Data Processing Automation

Postdoctoral Researcher

2017 - 2019
University of Freiburg
  • Led an analysis team for international collaboration to publish the world's most sensitive dark matter search results. The team had more than 60 scientists from 20 institutions around the world.
  • Headed a working group to design and install the data acquisition system for a more extensive detector upgrade. I managed a ten-person team and built and installed a €1 million system that is still in operation under an Italian mountain.
  • Performed data analyses across all sub-teams as needed to fill gaps in the workforce. I reviewed and challenged the results. Also, I drafted and wrote a seminal paper about the experiment, collaborating with 150 other people working on this project.
  • Taught at the university level. Advised graduate students both locally and in other teams. Presented at schools, conferences, and outreach events around the world.
Technologies: Python, C++, Statistics, Physics, Big Data, Electronics, Team Leadership, Remote Team Leadership, Node.js, Linux Server Administration, University Teaching

Postdoctoral Researcher

2013 - 2017
University of Bern
  • Designed and installed a data acquisition system for a large dark matter detector. I used a fast triggerless readout system, built structured events with No-SQL database queries, and ran 24/7/365 over the three-year lifetime of the experiment.
  • Led a detector physics analysis subgroup of about ten scientists. The main task was to study incoming data and ensure that any and all anomalies discovered could be reasonably explained and wouldn't interfere with the result.
  • Worked on the data formats, governance, processing systems, and analytical routines for what would eventually become 2PB of experimental data.
  • Taught, advised, and supervised graduate students. Also, I participated in outreach events and presentations at conferences worldwide.
Technologies: Python, C++, Django, Linux Server Administration, MongoDB, Electronics, Analog-to-digital Converters (ADC), Data Analysis, Software Development, Amazon Web Services (AWS), University Teaching, Physics

Machine Learning Marketing Platform

A machine learning production platform based on Kubernetes and PySpark. The data source for this project was an SQL and S3 data warehouse. The system performs a periodic processing pipeline, creating over a hundred tables (some based on each other) from the source data containing machine-learning-ready features. These features are trained into dozens of machine learning models and served to various consumers. Quality gates at several steps create automated alarms in case of exceptions and allow operators to view quality control metrics.

Data Acquisition System for Dark Matter Detector

https://arxiv.org/abs/1906.00819
A system that turns analog signals of a few megavolts on several hundred coaxial cables into structured data shipped off to cloud storage. The system was located under a mountain in Italy and operated all day and night for several years straight. It recorded data for what was considered the most sensitive dark matter detector in the world at the time.

Technologies included:

• Analog electronics, hardware signaling, i.e., a nuclear instrumental module (NIM logic), and field-programmable gate array (FPGA) routines
• C++ control and readout software parallelized over several physical servers and communicating over a unified network interface
• A MongoDB database layer with a distributed deployment over multiple servers
• Python top-level program logic
• A Django web service with the JavaScript front end
• A bare metal server cluster, local network, dozens of electronics components, and thousands of cables.

I built this system with a small group and was the project's leader. While I didn't implement every part myself, I did work on every part and was the main coordinator of the effort. Also, I was the official leader of the data acquisition team for a subsequent upgrade to this system.

Machine Learning Anti-money Laundering Platform

Used machine learning to reduce false positive money laundering alerts. The system runs automatically and live, preparing ML features using Apache Airflow DAGs and serving trained TensorFlow and XGBoost models over a live web API with only millisecond processing time.

Models are trained periodically and the system provides full traceability as well as model explainability reports, fulfilling compliance requirements.

This system provides a seven-figure cost savings with no loss in sensitivity.
2009 - 2012

PhD in Physics

Ruhr University Bochum - Bochum, Germany

2005 - 2009

Master's Degree in Physics

State University of New York at Albany - Albany, NY, United States

Libraries/APIs

PySpark, Node.js, TensorFlow, XGBoost

Tools

Jupyter, Emacs, GitLab CI/CD, Helm, Apache Airflow

Languages

C++, Python, JavaScript, SQL, Go

Platforms

Linux, Kubernetes, Visual Studio Code (VS Code), Amazon Web Services (AWS)

Storage

MongoDB, Amazon S3 (AWS S3)

Frameworks

Django, AngularJS, Spark, Flask

Industry Expertise

Project Management

Other

Data Analysis, Machine Learning, Physics, Mathematics, Cabling, University Teaching, Data Science, Data Analytics, Software Development, Big Data, Electronics, Monte Carlo Simulations, Linux Server Administration, Machine Learning Automation, Neural Networks, APIs, Data Processing Automation, Statistics, Team Leadership, Remote Team Leadership, Analog-to-digital Converters (ADC), CI/CD Pipelines

Collaboration That Works

How to Work with Toptal

Toptal matches you directly with global industry experts from our network in hours—not weeks or months.

1

Share your needs

Discuss your requirements and refine your scope in a call with a Toptal domain expert.
2

Choose your talent

Get a short list of expertly matched talent within 24 hours to review, interview, and choose from.
3

Start your risk-free talent trial

Work with your chosen talent on a trial basis for up to two weeks. Pay only if you decide to hire them.

Top talent is in high demand.

Start hiring