Enes Bol Email & Phone Number

Netherlands

İstanbul, Türkiye

• Churn Prediction Pipeline: Developed SQL data pipelines on GCP - BigQuery and deployed ML models using SQL, XGBoost, and Composer to generate daily churn risk scores.• Implemented CI/CD workflows for faster iteration cycles and reduced deployment time to 3 minutes. and increased overall system reliability by following best-practices.• Introduced best practices for repository management, documentation, and collaboration, improving code quality and accelerating development cycles. Guided the team members to ensure adherence to these practices, contributing to smoother project execution.• Planned project scope, timelines, and deliverables. Coordinated task distribution and ensured efficient progress. Provided technical guidance and clear roadmaps for successful project execution, aligning with business goals.• Building pipelines for data collection, governance, model training, and evaluation for Large Language Models. Developed robust pipelines using SQL and Python for data preparation, cleaning, and structured data extraction. Integrated and standardized multiple in-house and open-source data sources, and transformed large text databases into the required format for analysis and model development.

Gent, Flaman Bölgesi, Belçika

• Built a scalable data pipeline on Azure Databricks using PySpark to efficiently parse and aggregate billion-scale, nanosecond-level time series sensor data stored in Azure Blob Storage.• Built a data pipeline to parse incoming meter data into a list of time series, calculate statistics and metadata, and impute missing data for further analysis. Ensured data governance and quality while constructing tables to store forecast experiment results and compute key metrics like imbalance costs and profile costs.• Improved demand forecasting by incorporating covariate data, optimizing model parameters, and utilizing advanced boosting, tree-based, and transformer-based models alongside Prophet for time series analysis.• Developed robust backtesting, experiment tracking with MLflow, and hyperparameter optimization using Optuna for better model performance.• Built automated workflows to calculate hedging strategies, imbalance costs, and other key financial metrics, along with establishing API connections that streamlined data integration and improved the efficiency of analytics processes.

Londra, İngiltere, Birleşik Krallık

• Designed a data processing pipeline that leverages bert models to hierarchically categorizing extensive furniture data from100+ brands across 25+ retailers, resulting in improved search and user experience.• Implemented a robust data governance process, which included comprehensive data preprocessing with outlier detection,ensuring clean input for models and secure ingestion of resulting data into PostgreSQL.• Extracted critical information from product descriptions, enhancing the accuracy of product data.• Achieved increased website visibility, leading to Google recognition despite its recent launch.

Proficiently addressed a wide array of data challenges encompassing both structured and unstructured data. Performed data analysis, created Machine Learning and Deep Learning models, and developed ETL data pipelines. Demonstrated strong decision-making abilities and efficient project management, successfully handling multiple projects in parallel. Consistently delivered high-quality solutions, earning positive feedback from various business objectives.• Awarded "Top Rated" status on Upwork, reflecting exceptional client satisfaction and the ability to consistently deliver high-quality projects.• Streamlined House Property Data Extraction System: Developed a system for extracting structured house property details (e.g., room count, amenities) using the Llama-3-70B API from Together AI and Langchain-Kor. By reducing payload bloat and implementing custom field-by-field validation with Pydantic, I achieved a 75% reduction in token usage, a 50% decrease in delay times, and a 66% reduction in API costs, significantly improving performance and efficiency.• Built High-Accuracy Data Cleaning Framework: Built a framework to clean a chemical/biological/medical database of 2 million products by flagging typos in complex product names. Optimized the process to identify typos and compare them against ground truths in under 3 minutes, achieving +95% accuracy.• Keyword & Text Clustering Application: Implemented batch and parallel pipelines to retrieve OpenAI embeddings in under 9 seconds for 10k words, utilizing HDBSCAN and a custom-built pipeline for clustering logics to optimize clustering accuracy & functionality.• Enhanced lexicon coverage for a research paper by developing a Lexicon Extension algorithm using GloVe embeddings, aligning with NLP practices and contributing to an upcoming paper.• Automated structured data extraction from multipage PDFs using Amazon Textract and AWS Lambda, streamlining data handling and ensuring efficiency and scalability.

• Developed a CRUD API using C# in .NET with ASP.NET Core Web API and worked with PostgreSQL in a full-stack development environment, gaining valuable database experience.

İstanbul, Türkiye