Berke Can Rizai Email & Phone Number

Berlin, De

• Developed conversational LLM shopping assistant with fine-tuned Llama 2. Integrated API endpoints for product search and information retrieval. Developed orchestrator and agents with guidance and vLLM.• Built a recipe search engine for shopping assistant, leveraged DeBERTa to create a custom NER-tagged recipe dataset• Developed and deployed personalized product recommender models for the "Picked for You" page, boosting conversion rate by 62%, average basket value by 6%, and generating an estimated £350,000 monthly impact.• Built SOTA deep learning-based ranking models including DeepFM, Wide & Deep, Neural Collaborative Filtering, and Two Tower for personal homepage feed and deployed them as microservices.• Built an offline evaluation library for recommender models, accelerating iteration speed.• Established comprehensive integration with MLflow, enabling easy configuration of model parameters and features.• Tools: Python, PyTorch, TensorFlow, Lightning, FastAPI, AWS SageMaker, MongoDB, SQL, RedShift, S3, MLflow, Docker, Huggingface, Weaviate, LangChain

• Built a SOTA double machine learning (causal inference) pipeline to estimate price elasticity, enhancing pricing optimization and promotional strategies.• Developed optimization model for courier shift assignments.• Built a job recommender system leveraging BERT-based topic modeling, accelerated by NVIDIA Rapids for faster text processing.• Tools: Python, FastAPI, AWS SageMaker, MongoDB, SQL, RedShift, S3, Airflow, MLflow

• In the AI research team of Prof. Metin Sezgin.• Developed and integrated a handwriting recognition model using PyTorch for the PowerPoint AI project.• Added new features and improved user experience in a React.js web application.• Implemented unit tests for full functionality coverage, and created CI/CD pipeline.• Created MLOps pipeline for pre-processing drawing data, training and saving models, and logging results.• Tools: Python, Pytorch, Django, sklearn, JavaScript, React, S3, GitHub Actions

Istanbul, Turkey

• Created a web application with Java Spring Boot that retrieves and sends authentication codes to the sales force.• Built an ETL pipeline for transforming and processing unstructured data from over 1,200 clients.• Co-authored the pre-processing phase of an ML project for credit risk scoring.• Tools: Python, Java Spring, React Js., MongoDB, PostgreSQL

Istanbul, Turkey

• Created projects using Azure Cognitive Services (Vision, Text analytic, Search API, ML).• Completed courses on Azure solutions.

Istanbul, Turkey

• Developed a 2D Android platformer game using C# and Unity.

Bachelor Of Engineering - Be, Computer Engineering

Relevant Coursework: Machine Learning, Natural Language Processing, Data Privacy & Security, Applied Statistics, Optimization Methods.

Bachelor'S Degree, Industrial Engineering

Bachelor'S Degree, Computer Science

Relevant Coursework: AI Exchange student Received Study in Canada Scholarship

Lise