Saravana Kumar

Data Engineer with 4+ years of software engineering experience, including 3+ years building end-to-end GCP data pipelines across 4+ source systems. Proficient in BigQuery, PySpark, Apache Airflow, Dataproc, and Cloud Build CI/CD, with a track record of reducing BigQuery costs (~8%) through partitioning, clustering, and query optimization. Skilled in Python, SQL, and event-driven architecture, with experience supporting company-wide reporting via Looker Studio.

Data Engineer at Nublar Technology (2024-06 – Present)

Project: R8Connect - Logistics Rate Calculation Platform

• Architected GCP-based ETL/ELT pipelines integrating 4+ logistics and rate data sources into GCS, processing datasets with PySpark and selective Pandas-based transformations; engineered DataFrame-based workflows for cleansing, schema standardization, deduplication, and business-rule transformations before loading curated datasets into BigQuery.
• Refined BigQuery partitioning/clustering strategy for the event-driven pipeline built on Cloud Functions and Pub/Sub, reducing query processing cost by ~8% while automating processing of newly landed GCS files.
• Designed and operated 3-5 production Apache Airflow Directed Acyclic Graphs (DAGs) (GCS Sensors, Python/BigQuery Operators, XComs, retries, exponential backoff, failure callbacks) for reliable daily pipeline execution, and set up Cloud Build CI/CD pipelines to automate testing and deployment of DAGs and pipeline code.
• Managed Identity and Access Management (IAM) service accounts and least-privilege access controls across GCS, BigQuery, and Cloud Functions, and developed Looker Studio dashboards consumed for company-wide reporting.

Data Engineer at Marberx Technologies (2023-05 – 2024-03)

Project: Bible Application - Web & Mobile Data Platform

• Built distributed PySpark ETL pipelines on Google Cloud Dataproc to ingest and process structured CSV and JSON datasets from GCS, using the Spark DataFrame API for filtering, column standardization, aggregations, and business-rule processing.
• Provisioned an ephemeral Dataproc cluster (create, process, delete model) for PySpark workloads, lowering idle compute cost, with persistent datasets retained in GCS.
• Built efficient, partitioned Parquet datasets across raw and curated zones, producing 10+ curated datasets and tables integrated with BigQuery via the Spark BigQuery Connector for downstream consumption; performed data quality validation within Spark jobs and applied IAM least-privilege access controls across Dataproc and GCS.

Software Trainee at RND Softech Pvt Ltd (2021-11 – 2023-05)

• Developed and maintained an HR application for employee data, attendance, onboarding, and exit management using C#,.NET Core, Web API, and SQL Server.
• Engineered data annotation and preprocessing workflows for 500+ audio hours/month, delivering clean, structured datasets for Google AI speech recognition model training at 98%+ accuracy through rigorous QA and annotation review.

B.E. in Electronics and Instrumentation Engineering – Dr. Mahalingam College of Engineering and Technology (2017 – 2021)