Dorian Turba

Alpiq is major swiss producer and provider of electricity and gas for all Europe. Following the NIS2 UE directive, Alpiq needs support for upgrading practices across the SDLC and migrating business apps to the new standard. Missions: – Improve compliance to NIS2 and NIST CSF 2.0 for ~600 projects, – Lead the Pro-Code Center of Excellence Actions: – Develop, implement, deploy and maintain CICD Components for maximum reliability, security and maintainability for ~600 projects, – Create, promote and enforce Python projects new standards, – Sync with business and major stakeholders, document protocols and procedures to enforce relevant actions.

P1 Security is a Telecom CyberSecurity company that performs penetration tests for telecom operators and carriers. They have developed an automated system that report issues and vulnerabilities of the telecom network. Missions: – Implement CICD automation for projects, – Find and fix software security and reliability issues in projects, – Collaborate with developers to improve PenTests software. Actions: – Develop build, test and delivery CICD pipeline, – Audit and report vulnerabilities in Jira and implement fixes with developer training to prevent future defects, – Developed probes to test HLR and VLR vulnerabilities using SIGTRAN protocol and SS7 standard. Results: – Build stage under 15 seconds for both package and container build, test stage under 10 seconds for code quality and SAST, and operational image size < 88MB, – 3 probes done, highlighting one vulnerability and two potential ones.

Breitling is a Swiss luxury watchmaker known for its high-precision chronometers designed primarily for aviation use. They have a backend system for one of their services that was originally developed by an external company. Breitling needed an architect to design a new backend architecture to address technical debt and improve service scalability. Missions: – Audit software backend architecture, – Design and implement a new architecture with scaling capability as prime goal, – Ease adoption among current teams. Actions: – Audit and written reports of the current architecture, flaws and risks, – Concieved and documented a modular Clean Architecture Microservice-ready with state-of-the-art packaging, testing and release management, – Created mutltiple templates based on the previous design for each layer's modules, included automated tests, documentation and CICD. Results: – Each service, even when under high load, no longer impacts other services, – Each service and component is independently tested.

Sparks formation sells Python trainings from beginner to expert, and needs experts in the field in order to do them. – Following a predefined training plan, it is necessary to create course materials (theory and practice with micro-projects and exercises) and services in videoconferencing/on-site. – For each training, I have created a course support adapted to the needs of the clients, flexible on the topics, in order to maximize the satisfaction of the clients. – With the help of this support, each training is done in person on the customer's site or in videoconferencing. – Since the end of 2021, more than 50 professionals from the French Government, Siemens DIS, Sopra-Steria, Flowbird, and 5 other companies have been trained with an average satisfaction of 4.5/5.

Berger-Levrault acquired a Patient Record Management Software for healthcare facilities and required a strategy to transition from an on-premise software to a multi-tenant SaaS. I collaborated with a team that had been established for 15 years, including some junior developers, and faced numerous new security constraints and performance objectives. Missions: – Develop a Clean/Hex/Mach Architecture for Python, – Plan and adapt current software technologies and design, – Support other senior developers and lead juniors. Actions: – Create sample projects to discuss and collaborate on a new design to ease transition and on-boarding processes, – Write documentation to expose rational to the team and new members. Results: – Implementation of a three complete modular layers projects for HTTP-API, Logic and Persistence, – Write documentation to expose design rational to the team and new members, markdown as source of truth, exposed in Confluence, – Develop migration from Stateful API to Stateless API, Learnings: – Healthcare facilities business needs.

ECAM LaSalle is an engineering school. The digital engineering department trains their first-year students in algorithmics and programming. Missions: – Training of students on the basics of python (no OOP), – Transmission of good programming practices at beginner level, – Introduction to algorithmics. Actions: – Simple algorithmic exercises with the constraint of not being able to use almost all python tools and syntaxes, but having to recode them yourself, – Reading and lecture on basic and advanced Python concepts to prepare for the specificities of other languages (ex, call-by-reference/value/assignment, implicit/explicit cast, etc...), – Analyzing, processing, and notating code scripts by manipulating the abstract syntax tree (AST) of code written by students in multiprocessing. Results: – Over 300 first-year students trained.

RBI develop its own POS and BOH software for Tim Horton, Popeyes and Burger King brands in the US market and the international market. POS/BOH team is composed by ~250 people, ~65 software engineers, with a budget of ~50M$/Y. Missions: – Assisting projects CTO, – Drive technology and innovation of the project, – Take the technical leadership and expertise to reduce time to deliver of features and bugfixes and increase scalability and sustainability of projects, – Develop the team cohesion and create knowledge bridges between US and international teams. Actions: – Run senior candidate qualifications for projects, – Took the lead on workflow / technological / organizational shift, – Design architecture rework for kernel and localization of POS and BOH with the US tech leads supports, including automatic testing strategy and clean architecture guidance, – Manage cross-team communication and code-ownership exchanges from teams.

Saint-Gobain Glass France has several factories dedicated to glass plate production. In order to ensure that the finished products are of the highest quality, the company constantly monitors and adjusts the production process to optimize factors such as transparency, heat isolation, and color. This involves a significant amount of data analysis and computational work, which can be time-consuming. Missions: – Improve performances of algorithms that parse and deserialize optoplex sensors results files into Pandas Dataframes. Actions: – Audit of existing codes, – Development of new algorithms using different tools more efficient in context, – Developments of more robust tests to ensure resilience of the new algorithms. Results: – 500% performance gain on parsing and deserialization algorithms, – Cleaned code for parsing algorithms, following PEP20 and PEP8 guidelines and SOLID principles. – Learnings: – Better understanding of pandas multi-indexes, numpy arrays and pure python optimizations.

Hyperplan is a tool that helps agricultural cooperatives accurately forecast their local agricultural production. As the agricultural industry is constantly evolving, there is a high demand for reliable forecasting. Hyperplan was seeking a software engineer with expertise in Python, FastAPI, and SQLAlchemy to join their team and help them grow and become more stable and confident. Missions: – Improving the quality, maintainability and optimization of the code, – Transmission of general good practices to Python and specific to API development, – Improvement, design and development of the done and the to-be-done. Actions: – An audit of the existing situation with the goal of creating a book of best practices targeted on needs for future developments, – Design and implementation of a SOLID architecture allowing service dependency injections into the back-end in order to have a clear version management, simple to carry out and transmit to newcomers, while facilitating the creation of non-regression tests involving the database, – Participation in the development and optimization of the functionalities developed. Results: – +30% of the API is decoupled from its implementation, – +50% of the API behaves and is documented according to RESTFUL standards, – The size of tests has been reduced from about 50% to 80%, simplified implementation with +15% total coverage (60%->75%), – Tests can now be run in isolation with a run time (in isolation) increased from 15s to 25ms per test, – Contribution to the GeoJSON-Pydantic open-source project to improve the behavioural coherence of GeoJSON validators with Pydantic validators. Learnings: – Discovery of PostGIS, a PostgreSQL extension allowing the manipulation of geographic information, – Better understanding of Pydantic and GeoJSON-Pydantic data validation processes.

STM has created a software for their engineering teams to analyse and forecast microelectronic component defects. I was the Lead Python Backend & API in the API team and worked closely with other teams on this project. Missions: – Co-design and create the SQL Server database using Alembic, – Design and co-develop the API using FastAPI, – Ensure the non-regression of the API using PyTest, – Train a junior engineer on API and back-end server development. Actions: – Co-design of the SQL Server database model and creation of automatic migration scripts in close collaboration with data and SQL experts with Alembic, – Design and co-development of the REST SOLID API with FastAPI and SQLAlchemy, – Design and co-development of non-regression tests with PyTest on the API and database migration scripts. Results: – Efficient development and smooth coordination between the teams allowed 20 days spent on optimisation and quality of the API, – Over the last 40 days of the mission, partner was working independently on the development of functionalities, needing assistance only occasionally, – 100% automation of database deployment, – 85% code coverage, 100% of routes covered, 100% of database migrations covered. Learnings: – Multi-tenant databases, – Vertical and horizontal sharding in SQL, – Use of Alembic in the release testing processes, – First collaboration with Italians, Spaniards and Indians.

Senza Formations needs Python trainers to support its learners on Python and its environment. To do this, a monitoring of the progress of each learner as well as a verification of prior learning had to be put in place. Actions: – Exchange on the project and definition of a precise route record, – Validation of prior learning at each session, – Questions and exchanges on tools and good practices, – Small-projects and exercises to improve long-term knowledge.

G2S requires efficient performance from its PDF document analysis tools, as they can be slow to process. Additionally, there is a need to expand the types of data used for image analysis algorithms such as OpenCV and text detection. Missions: – Optimize PDF file processing processes (target: 50% improvements, first goal), – Fix a recurring bug on some pdf files (second goal), – Increase the size and diversity of the database used to train the image recognition algorithms (until the end of the mission, filler mission). Actions: – Performance tests on the code and benchmark of various tools available with cprofile, lineprofile and memprofile, – Code analysis and research to implement the necessary improvements, – Exchange with PyMuPDF development teams to determine the origin of a bug caused by the library, – Development of algorithms simulating software and hardware defects in cameras from existing documents with OpenCV and imgaug. Results: – Performance gains of 60-70% for PDF file processing, – Full bug patch with a temporary fix before release of the version implementing the patch in the library PyMuPDF, – 12 software and hardware defects simulated realistically in order to be digested by the algorithms. Learnings: – Better understanding of how OpenCV and PIL work, – Discover the imgaug library and its possibilities.

Flowbird parking meters require extensive development and maintenance. With installations in cities such as Lyon, Atlanta, and Jeddah, the team is working on a project with over one million lines of code in an international setting. Missions: – Contribute to the development of functionalities, – Maintain software and fix bugs. Actions: – Analysis of historical code, completion of automatic tests and simplifications of logical flows, – Creation of flowcharts in order to facilitate maintenance of the code that has undergone bug fixes, – Peer programming on the most critical bugs in order to accelerate the creation of the patch and share knowledge. Results: – 118n critical bug in Arabic corrected with 1 week of investigation and 1 week of fix in peer programming, biggest bug in mission, – Reduction of the code base by ~1k lines of code (~5k+, ~6k-) over the duration of the mission during simplifications. Learnings: – Training on electrical hazards while handling industrial machinery, – Deepening knowledge of lazyloading architectures particularly used in environments with constrained resources such as embedded computing, – 118n challenges.

EDF SF predict and optimize isolated or semi-isolated energy grids that rely on renewable energy production. EDF SF need to handle production forecast, consuption and storage, and flexibility. Missions: – Improve development velocity, maintainance and on-boarding processes, – Train their 15 engineers on software development on Python, – Diagnose and improve performances of algorithms, – Conduct code-review and ensure high-quality merges. Actions: – Create a company wide python SDK with automated quality-tests to harmonizing tools, improving maintainability, and facilitating update management. – Provided comprehensive training sessions on Python programming, covering OOP, coding standards (PEP8), and professional development tools (PyCharm, Git, etc.). – Conducted peer code-review of merge-requests, ensuring adherence to coding standards and best practices, while sharing insights to improve velocity with the team, – Profile and benchmark performance critical code, rework data pipeline to reduce network overhead, automated forward compatibility testing. Results: – SDK install reduced from hours of configurations to a single command line instruction, – Average of 80% less quality warnings, 30% less code-smells warnings per MR, – 10h long processes reduced to 6h, automated forward compatibility tested for 3 major upgrade of Python.