Melanie Aprahamian, Phd Email & Phone Number

San Francisco, CA, US

• Developed, prototyped, and built a streaming anomaly detection service on top of Flink that uses machine learning modeling to automatically detect anomalies for all golden metric time series (approx. 70 million time.
• Built an automated feedback loop that allows for continuous improvement of anomaly detection models by incorporating user feedback into the training process.
• Wrote and presented multiple internal blog posts and sets of documentation detailing both the high-level and technical workings of the anomaly detection service.
• Supported all of the services owned and maintained by the anomaly detection team by participating in on-call activities and responding to internal engineering questions.

US

• Modernized Nationwide’s enterprise model end-to-end deployment process by transitioning the advanced analytics organization off of on-premises tooling to AWS based cloud infrastructure with Amazon SageMaker and Kubeflow.
• Led and participated in the training effort for the entire data science community on the modernized model deployment infrastructure that was put into production at the end of 2021.
• Successfully converted five legacy machine learning models into a modern, cloud-based workflow.
• Defined and introduced the concepts of MLOps to the advanced analytics teams and how these concepts fit into the overall data and analytics ecosystem.
• Member of the Amazon SageMaker governance team that makes recommendations on best practices and usage patterns for the enterprise.
• Built and proctored a machine learning educational event for IT associates.

US

• Developed a Python method for automating a currently manual adjudication process for Pet Wellness claims.
• Developed method using PySpark and Databricks for supplementing telematics sensor speed data with speed calculations from GPS coordinates.
• Built a dashboard using Tableau to aid the Nationwide Discovery Management Unit Legal team.
• Built out data and machine learning centric educational workshops for both the internal Learnathon program and an AI/ML workshop for the Chief Technology Officer Cabinet.
• Completed a POC demonstrating the potential of utilizing MLflow with AWS native tools as a solution for machine learning model training, experimentation, and deployment.
• Participated in the vendor selection and tool modernization effort for the enterprise’s predictive modeling end-to-end solution.

US

Columbus, Ohio, US

• Developed methods to improve current protein structure prediction within the Rosetta software suite by introducing experimentally derived hydroxyl radical footprinting mass spectrometry data.
• Analyzed the Protein Data Bank for trends regarding the relative usefulness of each amino acid type for solvent exposure measurements to be used with protein structure prediction.
• Conducted extensive data analysis using Python.
• Developed a new Diffusion Monte Carlo (DMC) algorithm in Fortran using the rotational and torsional excited states of highly fluxional molecules like H5+.
• Developed C++ code that is included in the Rosetta software suite.
• Discovered positive inotropes for cardiac troponin C using computational small molecule docking software.

Columbus, Ohio, US

• Supervised undergraduate physical chemistry laboratory course involving both wet lab and computational experiments.
• Prepared and taught a weekly recitation session on chemical quantum mechanics and aided students with their assignments.

Doctor Of Philosophy - Phd, Physical Chemistry

Bachelor Of Science - Bs, Mathematics And Chemistry