Rezart Abazi

Developed a LangGraph-based agentic application for the Customer Care team as part of the organization’s AI transformation initiative. The agent integrated with Salesforce via APIs to retrieve and synthesize knowledge, performed complaint categorization, and delivered contextually accurate responses using a RAG pipeline with structured prompts and an LLM. Built on Azure Foundry and Databricks Vector Search with OpenAI embeddings, and integrated via a FastAPI service into the internal customer care tool to streamline operational workflows. Contributed aswell to the organization-wide AI strategy by defining key use cases and promoting scalable, business-aligned AI solutions.

Collaborated with the Data Science team, leveraging Python, to refine and update dynamic bidding machine learning algorithms for product positioning on major e-commerce platforms like Amazon, Walmart, and Target. Optimizing 2000+ campaigns and 20% cost savings for a similar or better share of shelf. Expanded existing data pipelines, integrating new retailers with SQL and Azure, to include data from our web-scraping provider API. Actively monitoring scheduled bidding and ETL pipeline jobs using Argo.Building end-to-end monitoring metrics using Prometheus (PromQL) and Grafana dashboards to assess the quality of scraped data and the reliability of our web scraping provider.

Building end-to-end processes with Python for downloading data in parquet format from SAP using ODAtaQuery with Prefect as data governor and orchestrated flows for data ingestion into Azure Data Lake. Additionally, implemented a non-clustered index using SQL.

Developed a solution to extract information from PDF-format documents using Azure Form Recognizer. Stored the extracted information in a tabular format in the database, utilizing Azure SQL for data operations.

Collaborated with the marketing team to design geospatial data models in SQL that substantially enhanced targeted marketing campaigns across Northern and Southern Europe, resulting in notable budget efficiencies for each campaign

Using python, I automated the transformation of Excel files, enhancing the data by appending column values. This streamlined process not only improved time efficiency but also simplified daily operations for the team, enabling easy handover.

Managed the settlement of mutual and hedge fund orders, and retrieved relevant data from the internal database utilizing SQL Server.

Co-operating with internal and external team stakeholders and clients to effectively deliver services and reports regarding their fund's activities.

Collaborating with the psychology department on a cognitive psychology research analysis using R-studio to analyze Facebook activities for a group of real students having as primary objective to build a prediction model able to predict gender based on the variables given by using the generalized linear model through the step-wise technique.

Some of the features determining the gender of a user were the number of messages, the amount of time being active on the problem, or the number of characters used for each message.