Serkan Coskun

• Used Knowledge Representation and Semantic Technologies (OWL, RDF, SWRL, SPARQL, JSON-LD), including semantic modeling and data integration, data unification, knowledge graph design, ontology, and taxonomy
• Created data/information architecture and modeling languages such as UML, ER, IDEF, Data Flow Diagram
• Worked with mapping techniques, R2RML, for transforming relational/tabular datasets into triples
• Developed end-to-end data pipelines with Python and used Cypher for transforming relational/tabular datasets into Neo4j
• Used py2neo and neo4j drivers for ingesting data from AWS Redshift and PostgreSQL into Neo4j DB.
• Used arrows.app to create Graph Models for communicating and explaining to the stakeholders

Used Python 3 and JupyterLab for developing use cases and ML models and used KubeFlow for the MLOps. Created data dictionaries, the ontology of the use cases, and Apache Atlas for data governance and metadata management. Developed business glossaries, specified the classifications, and added tags on the data. Used Apache Ranger for the central administration of security policies and monitoring of user access. Based on GDPR regulations, tagged PII (personally identifiable information) fields via Atlas and applied relevant data processing techniques such as tokenization/masking/anonymization to Ranger's appropriate fields/tags.

“AI Prediction Model and Root Cause Analysis” project that has been built using AI model and ML techniques to identify the key indicators of anomalies in manufacturing performance. Project benefits include reduced downtime, rectified breakdowns, predictable failures, allowed scheduling of maintenance activities, and control strategies for machine parameters to avoid breakdowns.

Worked on the Safety Stock Optimizer project. There is uncertainty in demand and in supply which may lead to shortfalls at the end of the replenishment cycles. Therefore, needs to be carried an appropriate amount of safety stock to protect against shortfalls. Safety Stock Optimizer project built-in AI/ML techniques to quantify demand and supply uncertainties and calculate a recommended safety stock level. This model created solutions for:
1. Demand uncertainty: uses forecast error calculation–a more realistic representation of actual planning behavior. Used decision tree to recommend optimum Safety Stock.
2. Supply uncertainty: variation of lead time against the average assumed to be representative of variation.
Used Azure Data Factory for end-to-end data pipelines. Made intensive use of delta functionality and combined batch and streaming workloads. Manipulated and transformed data in Microsoft Azure Data Lake Storage (ADLS) raw, foundation, trusted, and unified business logic layer. Used Azure Databricks for AI/ML model development.

Worked on the Smart Meter Transformation project and created/maintained optimal Data Lake and data pipeline architecture. Assembled large, complex data sets that meet functional/non-functional business requirements. Identified, designed, and implemented internal process improvements: automating manual processes, optimizing data delivery, re-designing infrastructure for greater scalability