Florian Elias

At Airbus: •Within the context of aircraft assembly, developed an NLP-based tool to categorize production anomalies based on their semantic meaning. •Optimized and restructured Spark data pipelines in line with MLOps best practices. •Developed machine learning models to accurately predict aircraft speed using only engine parameters, in case of main probe freezing. •Developed and deployed a solution to automatically detect faulty pressure sensors ( anomaly detection in spatio-temporal series) •Regularly presented progress and results to Airbus stakeholders.

•Tutored university students in Python / ML fundamentals. •Planned individualized lessons that address student’s needs. •Completed specialized online courses and certifications in areas such as MLOps and NLP.

•Participated in all aspects of machine learning projects aimed at moderating image and videos: data acquisition, annotation and cleaning,model training, evaluation, deployment and maintenance. •Implemented a facial recgonition model (facenet) for celebrity detection. •Trained smaller models on top of facenet (age ,gender prediction etc...). •Trained object detection models with Faster-RCNN ( Weapons detection, offensive content ,etc... ) •Explored multiple approaches for detection of hateful comments (BERT, LSTM, word2vec ).

•Filtering and cleaning of intraday stock data (high frequency time series). •Computation of risk indicators (VaR, Expected Shortfall, Beta). •Volatility estimation using GARCH. •Hierarchical clustering of all 30 Dow jones stocks with Dynamic Time Warping(DTW) and of all S&P 500 stocks with CORT.