Nelson Fonkoua

Vialink is a French LegalTech company specialized in the digitization and automation of KYC (Know Your Customer) processes. It had over 80 employees and supported major players in the banking and insurance sectors, with more than 20 million cases processed annually. Vialink was acquired in 2024 by Signaturit, a European specialist in electronic signatures and certified document management. I joined the "Identity" team as a Fullstack Engineer (Java / Angular), initially responsible for maintaining and improving the identity verification APIs and interfaces, as well as the authentication API used across all Vialink applications. My scope later expanded to include other KYC-related features. The team consisted - 2 Fullstack Developers - 1 Frontend Developer - 1 Product Owner - 1 Team Lead Key achievements: - Developed a configurable password expiration policy, which helped strengthen the overall security of user accounts. - Designed a REST API to automatically classify documents uploaded by users (e.g., payslips, tax notices). This automation reduced manual interventions and accelerated case processing. - I managed the progressive migration to a new identity verification API while maintaining full backward compatibility. This transition improved service reliability and incorporated new compliance standards. - Built a solvency evaluation API based on a customizable business rule language. It enabled automated solvency calculations using submitted payslips and banking information, streamlining user flows both on the front and back office sides. Technical context: Java 17, Java 21, Java 23, Kotlin, SCRUM, Agile Methodologies, Maven, Gradle, Linux, Spring Boot, Spring Webflux, Spring Security, Spring Data, Kubernetes, Hibernate, JIRA, Docker, code review, Clean Architecture, tests unitaires, Jenkins, Python, Angular, Git

Weborama is a French AdTech company founded in 1998, specialized in the collection, analysis, and use of data for advertising targeting and personalized marketing. With an international presence, the company offers data science and semantic targeting solutions for advertisers and publishers. In 2023, Weborama reported a revenue of €17.19 million and employed between 50 and 99 people. I joined Weborama as a backend Java developer. I initially worked in the BigSea team (4 backend developers, 1 team lead), responsible for the web user profiling platform. I then moved to the BigFish team (2 backend developers, 2 frontend developers, 1 product owner, 2 tech leads - backend and frontend), in charge of a web-based API for activating marketing segments. Key achievements: - Designed a high-performance CLI application to count unique users across hundreds of large files (several GB) stored on AWS S3. - Extended the BigSea platform to integrate a new user data provider. Implemented full data pipeline orchestration: ingestion, transformation into a unified format, and synchronization with existing services, significantly enhancing the quality and depth of generated web user profiles. - Developed a strategic feature for the BigFish API, enabling the creation and management of Contextual Segments based on keyword themes (e.g., sports, finance, travel). Implemented automated activation workflows and export of matching URLs to AWS S3, allowing clients to target relevant web pages and enabling new contextual marketing use cases. - Migrated user authentication to Keycloak while maintaining backward compatibility with legacy systems. Integrated Keycloak APIs for seamless user management and secure authentication flows. Technical context: Java 8, Java 11, Java 17, SCRUM, Maven, Gitlab CI, Linux, Spring Boot, Spring MVC, Spring Security, Spring Data, Kubernetes, Hibernate, JIRA, Docker, code review, Clean Architecture, unit tests

I worked on building a Machine Learning model that translates japanese sentences to french. The model was built with Fairseq It is hard to find a lot of ja-fr translations so I used OpenSubtitles's data but it was still not enough. So I had to generate more "fake data" and use some data augmentation method, one of which you can find on the provided link

Participant à l'implementation de la VPL (Verified Polyedra Library) Lien: https://github.com/VERIMAG-Polyhedra/VPL