Elijah Roussos
About Elijah Roussos
Elijah Roussos is the Lead ML Engineer at Cerebrium, where he has developed machine learning infrastructure and open-sourced dbt transformation models. He has a strong background in data science and engineering, with previous roles at Invictus Capital and the University of Cape Town.
Work at Cerebrium
Elijah Roussos serves as the Lead ML Engineer at Cerebrium since 2022. In this role, he has architected the Python package and infrastructure for the company's machine learning deployment platform, utilizing Amazon EKS and FastAPI. He has also contributed to the open-source community by developing dbt transformation models for SaaS data sources, which are now available as public dbt packages compatible with Airbyte. His work focuses on enhancing the efficiency and scalability of machine learning applications.
Education and Expertise
Elijah Roussos has a strong educational background in computer science and engineering. He earned a Bachelor of Science in Computer Science and Computer Games Development from the University of Cape Town, followed by a Bachelor of Science (Honours) in Computer Science in 2018. He furthered his studies at Cornell Tech, where he obtained a Master of Engineering in Electrical and Computer Engineering in 2021. Additionally, he completed a Data Science program at iXperience in 2019, enhancing his skills in data analysis and machine learning.
Background
Elijah Roussos began his career in the tech industry as a Web Development Intern at CloudAfrica in 2016. He later worked as a Data Science Intern and then as a Junior Data Scientist at Invictus Capital from 2019 to 2020. His experience includes tutoring 3rd-year computer science students at the University of Cape Town in 2018. His diverse background in data science and software development has equipped him with a comprehensive skill set applicable to various projects.
Achievements in Machine Learning
Elijah Roussos has developed a per product category inventory forecasting model utilizing XGBoost, neuralprophet, and GRU models. This model achieved test predictions within 2% of actual sales over time, demonstrating his capability in applying advanced machine learning techniques to real-world problems. He has also created a machine learning training pipeline using Prefect and Kubernetes, incorporating CI/CD with GitHub actions and enabling multi-core parallel training on an EKS cluster.