Yann Dupis

Leading the development and deployment of LLM-powered, multimodal content moderation systems across text, image, and video for top online platforms.

• Developed TF Encrypted (https://github.com/tf-encrypted/moose) and Moose frameworks (https://github.com/tf-encrypted/tf-encrypted) for encrypted ML (1K+ GitHub stars); co-authored paper at NeurIPS PPML 2018. • Fine-tuned and deployed Mistral 7B for PII detection (96% coverage) in secure, production-grade environments. • Built a confidential RAG system (Langchain, FAISS, OpenAI, AWS Nitro), used by 4K+ subscribers. •. Presented on privacy-preserving ML at TensorFlow World, ODSC, and Toronto ML conferences

• Lead a team of 10 data analytics specialists using exploratory data analysis, data visualizations, and statistical techniques to help auditors solve problems for more than 200 clients in the consumer and product industry • Oversee technical committees providing training and defining standards for R, SQL, SAS and Tableau, impacting the quality of 900+ deliverables

• Created applications in R/Shiny to improve the exploratory data analysis process on 900+ deliverables • Defined, prototyped, and scaled cognitive solutions (e.g. natural language generation, machine learning etc.) across the audit practice • Produced data visualizations using Tableau that improved auditors’ recommendations and efficiency by 50%. Results were presented to executives of major American companies • Teach R at new hire analytics Bootcamp and lead the R community in the Audit group.

• Led data driven research and analysis project on the impact of climate change and presented findings to key stakeholders. • Collaborated with marketing team to effectively use data analytics to study customer behavior • Member of the Intact Lab

• Using SAS and Excel, collected, compiled and analyzed all data requested by reinsurers to renew IFC reinsurance programs • Contributed actively to the discussions related to reinsurance programs renewal with reinsurance broker and different internal departments • Performed statistical analysis to generate key metrics to quantify earthquake exposure on portfolios based on catastrophe modeling software and reinforced various selected assumptions. Reported results to the executive management quarterly • Budget and allocate the reinsurance cost trough the company • Optimized, in collaboration with the underwriting department, the algorithm that calculates commercial property capacity • Reported commitment related to reinsurance requirement to regulatory authorities and AM Best

• Collaborated with the Oxford Internet Institute Research Team to explore how to promote social engagement in Massive Open Online Courses • Performed A/B testing to measure the impact of targeted email recommendations on social learning patterns • Presented key findings and insights to Oxford Internet Institute Research Team

• Parse web traffic logs and convert to have the data in a workable state • Analyze the data using Python to understand how early users interact with the platform • Present insights and recommendations using Tableau Software

• Developped software to manage the access of a wireless device CC1101 by a microcontroller ARM Cortex-3 • Developed in-depth understanding of a microcontroller ARM Cortex-3 and a wireless device CC1101 for use in design of communication between two wireless devices • Excelled as member of multi-disciplinary research team while sharpening C- programming and ARM skills

• Conducted market research on new surfboard technologies for wholesale retailer in sporting equipment industry • Identified new technologies and construction processes for inclusion in industry snapshot • Positioned product line in comparison with competitors; reported findings and recommendations to President and Commercial Director