Mohit Sharma

Programmer Analyst

Worked as member of internal product owner team (ML model deployment platform), my responsilibities include :

• Deploying machine learning models by converting python/hive codes to Spark
• Creating spark applications for data distribution and model performance monitoring
• Distributing model prediction step of keras based deep learning models using Pandas UDF
• Creating custom UDFs & JARs for use cases to be used in distributed data pre-processing
• Designing and management of internal facing products

SRM University

GPA: 9.23

Fraud Detection using GraphX

Developed spark application to help Fraud Detection for Apple using Spark GraphX. Providing fraudulent users data to train Machine Learning Fraud Detection Engine.

• Implemented unsupervised K-means clustering and fuzzy logic to enhance graphX result.
• Optimized the application performance by reducing data shuffle and optimizing joins.
• Utilized Pandas, Matplotlib and NetworkX libraries for visual representation of data and graphs.
• Automated data validation using Python.

ARES Semantic Process

Spark based semantic process for extracting key-value pairs from JSON data based on dynamic keys configuration in UI by user.

• Created hive tables and script to maintain data in hdfs.
• Data validation between hive source and target teradata database.
• Automated duplicate json keys validation using Python.
• Implemented Python script to drop hive tables partitions periodically.