Pankaj Kumar

Computer Science graduate (B.E., CGPA 8.33/10) with proficiency in Python, SQL, PySpark, and data modeling, and hands-on experience designing 2 end-to-end ETL/ELT pipelines and cloud-based data solutions across AWS, Microsoft Fabric, and Azure. Practical exposure to a 3-stage medallion architecture (Bronze, Silver, Gold), real-time data ingestion, version control (Git), containerization (Docker), and machine learning-driven analytics, including training a CNN-based intrusion detection model on 40,000 real-world automotive sensor records, gained through a completed Data Science internship and independent data engineering projects. Skilled at building scalable, production-style pipelines that transform raw data into analytics-ready datasets and support real-time decision-making.

Holds 3 industry certifications in SQL and Databricks. Immediately available for full-time Data Engineering or Data Analytics roles, seeking to apply strong problem-solving and cloud data skills to real-world business challenges.

Data Engineer Intern at Rubixe (2026-06 – Present)

• Gained hands-on experience with SQL, PySpark, Apache Airflow, Docker, Azure cloud, MongoDB, statistics, and BI tools (Power BI, Tableau), alongside delivering 2 PySpark notebook projects with written analysis reports.
• Snack Analysis: imported 2 nutrition datasets (Starbucks and McDonald’s) into PySpark, cleaned and aggregated them with PySpark and SQL, and compared nutritional attributes to determine the healthier of the 2 chains.
• Mall Customers Analysis: ingested the customer dataset into PySpark and applied statistical analysis to study customer behavior by spending score, reporting patterns and business insights.

Data Science Intern at VisionAstra EV Academy (2026-02 – 2026-06)

• Engineered a lightweight CNN-based CAN Bus Intrusion Detection System trained on 40,000 real-world CAN bus records (approximately 944 MB of raw data from a Hyundai Sonata vehicle) to classify 5 traffic classes in real time: Normal, DoS, Fuzzy, Gear Spoof, and RPM Spoof.
• Built end-to-end data processing and feature engineering pipelines to ingest and balance datasets across all 5 classes (8,000 rows each), applying a time-based 80/20 train-test split (32,000 training rows and 8,000 test rows) per class to prevent data leakage between sequential attack recordings.
• Developed and deployed Flask-based REST APIs and interactive dashboards to serve real-time inference results across all 5 threat classes, unifying detection output into a single monitoring interface while optimizing model efficiency for scalable, edge-based deployment.

B.E. in Computer Science – Sahyadri College of Engineering and Management, Mangaluru, Karnataka