Younès Adem

– Developing AI-driven outbound communication channels that help automotive dealerships increase sales and improve after-sales engagement. – Building deep integrations between AI and automotive systems to drive truly autonomous and action-oriented AI. – Extending the Sandra platform to integrate with outbound communication data, enabling seamless analysis and actionable insights.

– Engineered secure, scalable platforms to enable secondary use of personal data for research and innovation: – European level: Led development of a secure user portal enabling cross-border access to genomic and clinical data for researchers and healthcare stakeholders – National level: Designed and delivered a pseudonymization portal to support privacy-compliant, record-linkable data sharing across institutions – Impact: Improved data reusability and privacy compliance across institutions.

– Reengineered a mission-critical insurance risk calculation system, transforming a legacy monolith into a scalable, performant solution: – Refactored 30,000+ lines of code using design patterns and SOLID principles to reduce technical debt and improve maintainability – Built a robust testing framework with 90%+ code coverage with unit and integration tests, significantly reducing production defects – Implemented CI/CD pipeline with automated quality gates, testing, and deployment to ensure reliable and iterative releases – Delivered full-stack solution with Python backend calculations and React frontend for real-time risk calculations and regulatory reporting – Impact: Reduced calculation processing time by 60%, eliminated manual deployment risks, and created flexible architecture ready for future regulatory changes.

– Designed end-to-end machine learning solutions and scalable data infrastructure to drive operational efficiency: – Built scalable data pipelines with complex ETL transformations using Python and Azure Data Factory – Developed demand forecasting models with XGBoost, achieving 89.9% error reduction through advanced feature engineering and hyperparameter tuning – Automated contract classification through Python algorithms integrated with Tableau dashboards, eliminating manual bottlenecks – Delivered actionable insights via statistical analysis and interactive dashboards for stakeholders across business units – Impact: Transformed demand forecasting accuracy and operational efficiency, enabling proactive supply chain management and improved customer satisfaction.

– Designed and developed a Speech-to-Text system based on LLMs: – Benchmarked cutting-edge models (RNN-CTC, Seq2Seq, Transformers, Wav2Vec) – Trained and fine-tuned Wav2Vec on Mozilla Common Voice using transfer learning – Designed and implemented full production-ready pipeline for model inference and deployment, including comprehensive data processing for audio (resampling, MFCCs) and text (tokenization, normalization) – Built scalable data workflows (parallelization, prefetching, caching) – Impact: Enabled the startup to accelerate development of its AI assistant by delivering a scalable, high-accuracy Speech-to-Text system tailored for voice interaction.

– Enhanced a pedestrian counting system using advanced object detection and tracking techniques: – Researched state-of-the-art detection & tracking algorithms (YOLOv5, Deep SORT) – Designed custom evaluation metrics tailored for real-world pedestrian counting – Retrained YOLOv5 for improved detection accuracy in dynamic environments – Fine-tuned Deep SORT to boost multi-object tracking performance and reduce ID switches – Impact: Reduced counting error by over 90.61%, significantly outperforming the baseline model and enhancing overall system accuracy.

– Built internal web applications to streamline agile workflows and boost team collaboration: – Developed a real-time notification service to keep teams updated on project milestones – Created a project dashboard with live status tracking and notifications for timely follow-ups – Built a poker planning tool to support remote sprint estimation sessions – Impact: Enhanced project visibility and improved collaboration across teams, accelerating agile processes.