Mahmud Nabi

Migration project to upgrade Optimizely's data warehouse to a modern cloud system (Snowflake), reducing cost and data reliability across the organization.

• Migrating Airflow ELT pipelines to Argo Workflows on Kubernetes with dbt, enabling retries and backfills to achieve a 99.7% pipeline success rate
• Deployed Airbyte on AWS EKS to replace Fivetran, cutting cloud costs by 50%
• Designed Medallion architecture-based DBT pipelines enabling analytics across 30+ mission-critical business dashboards
• Implemented Point-In-Time tables in DBT by leveraging DBT macros, eliminating ambiguity across different but relevant data by 100%
• Implemented Terraform-based Snowflake RBAC for 2,000+ tables, enabling data security and governance with auditable access control

AI-powered data query engine for general knowledge discovery in Optimizely's data.

• Deployed a knowledge-graph-based AI agent for semantic data discovery, serving 40% of monthly recurring adhoc data requests
• Built an Actor-Critique AI agent (using Google ADK) for automated Snowflake SQL generation, reducing manual query effort by 35-40%
• Implemented agent-driven, schema-aware data quality tests with DBT, Argo, and GitHub Actions, reducing production incidents by 30%

Seamless integration of Salesforce and Snowflake, syncing customer insights from Snowflake to Salesforce to power smarter targeting.

• Engineered a high-throughput Python-based Snowflake-Salesforce CRM integration, reliably ingesting up to 30 million rows per day
• Refactored the legacy pipeline to incremental, idempotent processing, achieving atleast 50% reduction in cloud run costs

Scalable ML pipeline infrastructure, reducing deployment time and accelerating Optimizely's ML-driven customer churn prediction analytics.

• Optimized ML retraining by offloading pandas-based data transformations to Snowflake SQL, cutting runtime 60%
• Scaled churn prediction analytics 5x by orchestrating 20+ ensemble ML models with memory-efficient, fault-tolerant Argo pipelines on AWS EKS

Implemented PySpark data ingestion jobs for high-volume data processing.

• Implemented PySpark data ingestion jobs (up to 30 GB/day from MySQL and MongoDB), optimizing joins and storage
• Delivered analytics that cut go-to-market strategy review time by 3 days