Resource Modeling Solutions (RMS) develops an industry standard geostatictical python package (rmsp) and platform with state of the art functionality.
The task of developing the data conversion pipeline and in-notebook 3D Viewer was entrusted to me. Our goal was to present complex data in an intuitive and interactive format for our clients. The viewer builds upon vtk.js for interactive viewing.
Utilizing cloud functions, a large language model equipped with Retrieval Augmented Generation (RAG) capabilities was integrated to facilitate coding and documentation searches for users. The implementation leveraged RAG to ensure that the language model's responses were grounded in real-time and proprietary data, thereby providing accurate and relevant assistance.
Other projects included integrating third-party technologies to enhance product offerings. One notable achievement was the integration RMS's DrillHole Optimizer C library into rmsp or prototyping the integration between rmsp and GeologicAI's Coretable.
Climatematch Academy, a globally accessible summer school program, aims to democratize access to computational methods in climate science. It provides a comprehensive introduction to these techniques, fostering a diverse community of learners committed to addressing climate challenges.
As the tech lead, I oversaw the development of DevOps pipelines (using GitHub Actions) to facilitate code testing and the creation of a Jupyter Book.
Another of my roles was to curate a unified environment (based on the pangeo stack) that would work with all course material and provide students with all the tools they needed.
Finally I was the point of contact with 2i2c who setup and maintained a JupyterHub for the course.
CEREGE (Centre de Recherche et d'Enseignement des Géosciences de l'Environnement) is comprised of several specialized units, one of which is the climate team. Within this team, I was actively involved in the paleoclimatology workgroup, collaborating closely with researchers across France and Europe.
Configuring boundary conditions for climate models, such as those provided by IPSL, can be intricate. To enhance usability, I developed, deployed and maintained a Software as a Service (SaaS) platform that enables users to construct the necessary inputs interactively. The project was open sourced on GitHub.
I contributed to the development of a semi-automated processing pipeline for analyzing results, which was successfully deployed on an internal JupyterHub instance that I managed. This deployment facilitated efficient and streamlined processing of data.
During my time at CEREGE I engaged with the Pangeo group to enrich my understanding of computational climate science and its best practices.
Envisol is a French Consultancy specialized in Contamination and Remediation.
During my stint at Envisol, I was entrusted with the responsibility of constructing a prototype geostatistical Software as a Service (SaaS) solution. This project aimed to accurately characterize and regularly update contamination estimates. Leveraging advanced geostatistical techniques, the solution was designed to provide detailed insights into contamination levels, supporting informed decision-making and remediation strategies.
Further work included automating data acquisition workflows and building bespoke tools for the QGIS platform.
Seequent is recognized for creating the Leapfrog 3D suite, a leading solution in the geological modeling sector. The software uses Radial Basis Functions (RBFs) to generate surfaces in an implicit manner from sparse input data.
As part of the R&D team, I collaborated with various internal teams and clients, offering technical expertise and producing prototypes. These prototypes were either developed by repurposing existing Core IP or by exploring innovative approaches, such as leveraging cloud-based computations and web dashboards. This multifaceted approach allowed us to rapidly prototype and gather valuable user feedback, iterating on our designs to refine and enhance our solutions. By integrating serverless technologies from major cloud providers, we were able to build and deploy prototypes efficiently. This methodology enabled us to deliver high-quality, user-centric solutions.
Ecole Nationale Superieure de Geologie is a leading French engineering school specializing in geo- sciences and delivering an Engineering diploma combined with a Master from the University of Lorraine.
Title of Thesis: ”Current automatic well log correlation techniques, their advantages and drawbacks”.
I am at ease in swiftly crafting tools and workflows to explore new ideas and iteratively refining them in collaboration with stakeholders. By leveraging rapid prototyping and iterative design processes, I've been able to transform initial concepts into tangible, user-focused solutions.
I've been fortunate enough to work in development teams across different company stages and sizes. This experience has enriched my understanding of software development and coding standards, teaching me not just what these practices entail but also when and how to effectively implement them.
Having worked across multiple scientific computing companies, I have gained extensive experience in writing, debugging, and optimizing code in various programming languages. This versatility allows me to effectively interface between different languages, leveraging the strengths of each to achieve optimal performance and efficiency.
Proposing code and data objects to the web or a Software as a Service (SaaS) platform through API creation is crucial. While I am acquainted with various best practices and frameworks, my expertise lies particularly within Python's Flask and FastAPI.
I've been involved in various projects ranging from developing Flask-based applications for local deployment to orchestrating intricate processes across numerous containers. Additionally, I've explored the application of serverless functions for engaging with Large Language Model (LLM) inference APIs and integrating cloud-stored data.
Being able to visualize data intuitively and interactively is primordial to being able to rapidly iterate on ideas. I keep up to date on the latest advances in the domain.
Transitioning to the cloud presents a myriad of advantages and disadvantages. Cloud architectures inherently necessitate a microservice architecture, which requires varying degrees of granularity. Throughout my career, I have had the privilege of designing and orchestrating such architectures on multiple occasions. This experience has underscored the importance of implementing best practices in microservice orchestration, such as monitoring and logging services and designing for failure to ensure fault tolerance.
I am familiar with using and building using containerization technology as well as using serverless runtimes.
To effectively communicate complex scientific knowledge and algorithms it is essential to have a frontend. In small agile teams lacking dedicated frontend developers, it's crucial to adopt a combination of best practices in frontend development and leverage backend expertise for effective products.
Coauthor of JupyterLab quick start guide. Available now on Github.
Presentation in front off over 100 Oil and Gas experts demonstrating my Master’s thesis on automated well log correlation techniques. The Master’s thesis was also published in the Proceedings.
Presentation demonstrating the integrations between Building Information Modeling and subsurface data.
Short talk on building a infrastructure to rapidly build and deploy prototypes.
Presentation in front of over 200 mining experts on behalf of Seequent demonstrating the latest R&D work carried out at the company.
Build a tool to design filters for seismic data in the frequency domain. After the hackathon a web based tool (panel) was built and deployed to AWS.
Refresh of the website to show boat rescues in the mediterranean sea.
Take part in weekly catchups to discuss and better understand the intersection between computing and climate sciences
Build a web tool (Dash / plotly) to load LAS files and visualize them.
SaaS web app to setup boundary conditions for the IPSL model.