At AON my goal is to build AI for extrusion based processing of high performance polymers.As an individual contributor, I conduct novel research and build prototypes to validate research hypothesis using applied math and modern AI approaches for simulating physical phenomena.As lead of the R&D team, I am responsible for setting and pursuing research goals for an interdisciplinary team covering polymer scientists, mechanical engineers and software engineers. Collaboration and shared understanding are key to success in interdisciplinary teams.

At Vathys, I worked on researching and prototyping approaches that lead to improved post-training neural network performance. Using a variety of techniques, a few of which became standard in recent years: quantization, trimming low activation weights, on-the-fly tensor operation rewriting using compiler theory, convolutional kernel operation dimensional decomposition.

As Director of Engineering I was in charge of the company technical vision, technical department management / operations and staffing growth of our technical department with a global recruitment outreach, focusing on Western and Eastern Europe. Over the course of a year I was able to more than double the size of the engineering department, going from 14 engineers to 32. In order to support the large department growth I set up a new department structure with both junior technical management recruited internally and more experienced technical management from external sources. This resulted in growth opportunities for existing employees and a clearer path forward through coaching from experienced practitioners.

Currently I am leading the R&D and data teams at 3D Hubs.As the leading platform for online manufacturing in the world, we see a large number of 3D designs uploaded every day. It is my responsibility to make sure that we automatically analyze and quote them. Being on the critical path to client conversion, it is imperative that our analysis platform have performance as close to real time as possible. We use a mix of C++, Rust and scientific Python for the number crunching geometric analysis and another mix of Keras and Tensorflow for building the decisional side of our platform. My main contribution is designing and developing the performance optimized manufacturability analysis, with a median client request waiting time of 21 seconds and a 90-percentile of 128 seconds, representing a 15x improvement over our competitors. Once released, an A/B test revealed an 11% AoV improvement and a decrease in order price changes from 35% to 3%.As lead of the data team I work closely with our business analyst. My main contributions are in formulating business hypotheses in terms of statistics that I then implement as analytics on top of our data framework. Tools of the trade include SQL, AWS Redshift, Airflow for automation and orchestration, and finally Python for pipelines and statistics.

Printivate was a 3D model repair and optimization platform for 3D printing, exposed asa 3D model processing SaaS platform that was easy to integrate into existing products through a flexible API.I developed Printivate in C++ (backend processing platform), Clojure (web framework) and ZeroMQ.In 2016 Printivate was acquired by 3D Hubs.

Ullink is a world leading electronic trading solutions provider. In a small agile team, I am working on the risk management component, a backend Java module with stringent performance requirements in terms of response latency and throughput.With SLA latencies in the order of 1ms, my role involved implementing new types of risk mitigation strategies. As a separate individual contribution, I analyzed and formalized all the different types of risk analysis present in the component and used this analysis to redesign the component with performance in mind.

XtremLogic is a spin-off company that grew out of a research project from the Ecole Normale Supérieure de Lyon, France. The goal is to produce high-performance integrated circuit designs in the form of Intellectual Property blocks.As part of a small team, I worked on the design and development of the high level synthesis framework that generates circuit designs from high-level programs written in C. The framework itself is based on cutting-edge research in the fields of compilation, high-level synthesis and high-performance computing of the French institute INRIA.

As part of the distributed systems and parallel algorithms laboratory, I followed a PhD program on the topic of Resource management for large scale Cloud Platforms at Ecole Normale Supérieure de Lyon in the Avalon research team, under the supervision of Frédéric Desprez and Eddy Caron. Over the period of my PhD I worked on automatic application scaling and criterion-based deployment of large-scale applications on Infrastructure as a Service Cloud platforms.

As a Microsoft Student Partner (MSP) I was responsible for learning and presenting the state-of-the-art of development languages and tools developed by Microsoft.