Younès Adem

Built an AI-powered customer engagement platform for automotive dealerships, enabling autonomous multi-channel outreach and agentic workflows: • Designed and implemented a multi-channel conversation system handling SMS, RCS, WhatsApp, and Email — including smart follow-up logic to drive autonomous, context-aware customer interactions • Integrated LLMs with real agentic capabilities by connecting models to dealership DMS via MCP and custom tools, enabling AI agents to take meaningful actions within dealer operational systems • Built post-processing and analytics pipelines to extract insights from generated conversations, powering real-time performance visibility • Developed an email notification system surfacing key conversation events and summaries to dealership teams • Delivered a SaaS analytics platform giving dealerships visibility into campaign performance and individual conversation outcomes Impact: Enabled automotive dealerships to run fully automated, multi-channel customer engagement campaigns with end-to-end observability — from outreach to conversation to actionable analytics. In 9 months, the platform processed 90,000+ contacts, achieved an average 20% response rate, and drove 5,000+ booked appointments. Skills: Python · FastAPI · TypeScript · React.js · Docker · Terraform · Amazon Web Services (AWS)

Engineered secure, scalable platforms enabling privacy-compliant access to sensitive health and research data across national and European ecosystems: • Led the design and development of a secure cross-border data access portal, enabling researchers to access genomic and clinical datasets in compliance with European regulations • Designed and implemented a pseudonymization platform supporting record linkage across institutions while preserving strict privacy guarantees • Built and secured authentication and authorization flows using identity management systems, ensuring fine-grained access control for sensitive data • Developed scalable backend services and APIs to handle data access workflows, auditability, and interoperability between heterogeneous systems • Collaborated with cross-functional stakeholders (legal, data governance, healthcare experts) to translate regulatory constraints into robust technical solutions Impact: Enabled secure data sharing across institutions, improving data reusability while ensuring compliance with privacy regulations. Skills: TypeScript · Node.js · React.js · Docker

Re-architected a mission-critical insurance risk calculation platform, transforming a legacy monolith into a scalable, testable, and production-ready system: • Led the refactoring of a 30,000+ line legacy codebase, applying design patterns and SOLID principles to significantly reduce technical debt and improve maintainability • Engineered significant performance gains by optimizing core calculation algorithms, introducing caching layers, async processing, and targeted query/database optimizations — cutting end-to-end processing time by 60% • Designed and implemented a comprehensive testing strategy (unit, integration), achieving 90%+ coverage and drastically reducing production defects • Built and deployed a CI/CD pipeline with automated quality gates, ensuring reliable, repeatable releases and faster delivery cycles • Developed a full-stack solution combining Python-based risk calculations with a user interface for real-time regulatory reporting • Collaborated with business and regulatory stakeholders to translate evolving requirements into a flexible and extensible system architecture Impact: Reduced calculation processing time by 60%, eliminated manual deployment risks, and delivered a system adaptable to future regulatory changes. Skills: Python · FastAPI · TypeScript · React.js · Docker · Microsoft Azure

Designed and deployed end-to-end machine learning solutions and data infrastructure to optimize industrial operations and supply chain efficiency: • Developed and productionized demand forecasting models using XGBoost, achieving ~90% error reduction through advanced feature engineering and hyperparameter optimization • Built scalable data pipelines with complex ETL transformations using Python and Azure Data Factory, enabling reliable processing of large industrial datasets • Automated contract classification workflows using Python, integrating results into Tableau dashboards to eliminate manual processing bottlenecks • Delivered data-driven insights through statistical analysis and interactive dashboards, supporting decision-making across multiple business units • Collaborated with domain experts to translate industrial constraints into robust, deployable data solutions Impact: Significantly improved forecasting accuracy and operational efficiency, enabling more proactive supply chain planning and better customer service. Skills: Python · FastAPI · Microsoft Azure

Designed and deployed a scalable speech-to-text system to power voice interactions for an AI assistant: • Evaluated and benchmarked state-of-the-art ASR architectures (RNN-CTC, Seq2Seq, Transformers, Wav2Vec) to identify the best trade-offs between accuracy, latency, and scalability • Fine-tuned Wav2Vec models on Mozilla Common Voice using transfer learning, significantly improving transcription accuracy on target use cases • Designed and implemented an end-to-end inference pipeline, including audio preprocessing (resampling, feature extraction) and text normalization, ensuring production readiness • Built efficient data processing workflows (parallelization, prefetching, caching) to optimize training and inference performance at scale • Contributed to integrating the model into downstream systems enabling real-time voice interaction Impact: Delivered a high-accuracy, production-ready speech-to-text system, accelerating development of the startup’s AI assistant and enabling scalable voice-based interactions. Skills: Python · FastAPI · PyTorch · Docker · Terraform · Amazon Web Services (AWS)

Improved a real-time pedestrian counting system using advanced object detection and multi-object tracking techniques: • Evaluated and benchmarked state-of-the-art detection and tracking models (YOLOv5, Deep SORT) to identify optimal approaches for real-world deployment • Retrained and optimized YOLOv5 for robust pedestrian detection in dynamic environments (occlusions, varying lighting, crowd density) • Fine-tuned Deep SORT to improve tracking consistency and significantly reduce ID switches in multi-object scenarios • Designed custom evaluation metrics tailored to pedestrian counting accuracy, better reflecting real-world performance than standard benchmarks • Contributed to integrating detection and tracking components into a cohesive pipeline for reliable end-to-end counting Impact: Reduced counting error by 90.61%, significantly outperforming the baseline and improving system reliability in production-like conditions. Skills: Python · PyTorch

Designed and built internal web applications to improve team productivity and streamline agile development workflows: • Developed a real-time notification service to keep teams aligned on project milestones and updates • Built an interactive project dashboard with live status tracking and automated notifications to improve visibility and follow-up • Designed and implemented a planning poker tool to support distributed sprint estimation and remote collaboration Impact: Increased project visibility and team alignment, enabling more efficient collaboration and smoother agile execution. Used by 3 teams across multiple projects. Skills: TypeScript · Node.js · React.js · Amazon Web Services (AWS)