Arda Kosar

• Leading a growing team of 4 data professionals, including data scientist, data engineer and a software engineer, to deliver high-impact internal and client-facing projects.
  
• Built two internal Retrieval-Augmented Generation (RAG) agents to support internal teams in accelerating the transition from raw data to actionable insights. Leveraged Azure AI Search as the vector database for efficient and scalable retrieval, LangGraph for agent orchestration, and LangSmith for robust monitoring and evaluation. These tools significantly reduced manual analysis time and improved operational efficiency across key functions.
  
• Implemented a deep research agent capable of autonomously conducting multi-step research tasks and delivering results in user-friendly formats. Integrated features to generate and email structured reports as PDFs and produce audio summaries in podcast form, streamlining knowledge consumption for internal teams and enhancing accessibility to insights.
  
• Implemented a backlog and ticketing system in Azure DevOps to streamline project development and sprint planning. Introduced the Scrum framework to structure workflows, prioritize deliverables, and improve visibility and alignment across the team.

• Hired and managed a software engineer, providing technical guidance and project oversight while fostering a culture of ownership and delivery.
  
• Led the design and development of an internal SaaS platform to productionize proprietary data tools, enabling self-serve access for other teams and eliminating the bottleneck of manual support. This significantly improved operational efficiency and sped up the data to insights timeline.
  
• Developed a logistic regression-based propensity model in BigQuery ML to predict the likelihood of users clicking a key CTA on a website, using behavioral data from Google Analytics. Engineered session-level and user-level features, integrated model insights into reporting tools for product and marketing teams, and drove UX optimizations that resulted in a 150%+ increase in digital retail initiations.
  
• Developed a customer segmentation model using K-Means clustering on CRM data to identify distinct user groups based on behavioral and transactional attributes. This segmentation enabled the marketing and CRM teams to deliver targeted, user-level strategies—such as personalized offers and messaging—resulting in more relevant customer engagement and improved campaign performance.
  
• Designed and implemented a high-throughput data pipeline to process over 10K daily AVRO files from client's CDP to Google Cloud, utilizing Pub/Sub for messaging, Cloud Scheduler for automation, and Cloud Functions for processing. The pipeline wrote data to BigQuery, powering a live dashboard that enabled real-time reporting and analytics for business stakeholders.

• Conducted recurring analysis of 30M+ transactions for direct mail campaigns using Azure Databricks and Pyspark; delivered actionable insights that shaped coupon strategy and creative for future flyers, influencing millions in promotional spend.
  
• Developed customer lifetime value (CLTV) models to segment customers into actionable cohorts; enabled tailored messaging strategies that improved personalization and campaign relevance.
  
• Maintained and improved proprietary internal tools that enabled strategy teams to develop data-driven business strategies, enhancing decision quality and analytical self-sufficiency.

• Managed and improved Data Science teams CI/CD pipeline for data scraping and storage. Used Azure kubernetes-as-a-service for web scraping in parallel which decreased time to completion from ~5 days to ~1 day

• Used Python to automate processing and merging 1000s of csv files. Decreased 8-20 hours/week of unnecessary labor for teammates

• Deployed a multi-label classification model for classifying long-form texts, using scikit-learn, achieved %70 accuracy among 1400 paragraphs

• Created a methodology and designed the algorithm for identifying and mapping spheres of influence within social media networks which generated more than $500k/year sales

• Built interactive web application, using R and the Shiny web framework, to integrate behavioral science with Data Science for strategic and creative testing which generated more than $200k/year.

• Analyzed MySQL data with R and used ggplot visualization framework to deliver insightful visualizations to stakeholders.