Yasmine Chelly

• • Developed a music metadata matching pipeline with a human in the loop and a callback trigger that blacklists mismatches using google workflows, cloud Functions, Firestore and BigQuery. (Matching Project)

• • Collected and standardized streaming, TV, and radio data from multiple sources. (Enrichment Project)

• • Processing of client documents in various formats, with extraction of works and revenues to ensure data reliability and centralization. (Client Documents Project)

• • Developed API endpoints that implemented search functionality on the Claimy platform using Elasticsearch. . (Claimy Platform Backend)

• • Calculated artist earnings from collected data by considering multiple factors—such as Spotify stream counts and, for TV and radio, program duration, type, and time slot—and compared the results with the artist's actual earnings to identify missing royalties. Compare Project)

• • Helped on extending the Compare project to estimate future royalties, by revenue forecasting for artists through predictive models. (Estimations Project)

• • Generating of claiming proofs for artists to recover missing royalties, with automated data processing and calculation workflows. (Claiming Proofs Project) Technologies: .Golang, Python, Gitlab, BigQuery, firebase, elastic search, influxBD, Neo4j, Google Cloud Storage, cloud functions, Google workflows and schedulers, Dataflow, Pub/Sub messaging, Application Integration

• • Built data pipelines to extract, transform, clean and load big volumes of data using python, SQL and pyspark.

• • Orchestrated and monitored 40+ pipelines using Airflow.

• • Helped in recruiting and mentoring developers by organizing meetings and reviewing code in a team of 20+ data engineers.

• • Participated in defining and implementing MLOPS and DataOps pipelines using different open source tools such as Grafana, airflow, MLflow, etc.

• • Trained and deployed a Linear regression model using xgboost to estimate the media budget used in running ads.

• • Trained a forecasting model using Facebook's Prophet to predict required supplies for each store and region. Technologies: Python, SQL, Spark, Kafka, Airflow, cassandra, postgresql, Superset, Pycharm, Terminal, AWS S3, Bitbucket et gitlab

• • Developed machine learning models (NER, encoder/decoder models) that process invoices and checks and categorizes different fields.

• • Benchmarked different solutions to extract structured data from semi structured documents.

• • Implemented a Proof Of Concept that detects tables from PDFs. Technologies: Python, Pandas, Numpy, Tesseract, SpaCy, LayoutLM, encoders/ decoders.