At a cloud hosting provider, I built 50+ production dbt models in ClickHouse and optimized them with partitioning, LowCardinality columns, and scheduled partition merges, cutting CPU time by up to 50%. I also automated weekly warehouse checks with Airflow, freeing up to 50% of storage and reducing average RAM usage by 30%.
I built and deployed text classification and summarization services on AWS EKS, reaching 96% classification accuracy. I also developed feature pipelines served through Feast, production Airflow DAGs, and a CI/CD pipeline with GitHub Actions and ArgoCD; my work includes monitoring the classifiers with Prometheus and Grafana.

