Ömer Özgür Email & Phone Number

1-Designed, developed, and implemented generative AI models using state-of-the-art techniques, focusing on the creation of superagents for financial and energy sector applications.2-Created a financial data extraction tool using the OpenAI API, capable of extracting key financial information from news articles, providing valuable resources for financial analysts, researchers, and data scientists3-Developed an OpenAI API wrapper for handling consecutive calls and errors, ensuring robust… Show more 1-Designed, developed, and implemented generative AI models using state-of-the-art techniques, focusing on the creation of superagents for financial and energy sector applications.2-Created a financial data extraction tool using the OpenAI API, capable of extracting key financial information from news articles, providing valuable resources for financial analysts, researchers, and data scientists3-Developed an OpenAI API wrapper for handling consecutive calls and errors, ensuring robust and error-free code Show less

1.Developed and implemented a complete end-to-end data pipeline in AWS using SageMaker and AWS Batch, including data ingestion, preprocessing, model training, and model registry.2.Combined U-Net with Transformers to create a cohesive system for segmenting and classifying elements in Retinal CT images.3. Deployed pretrained models on AWS Sagemaker, AWS Lambda, Docker, Kubernetes.4.Developed and deployed deep learning models for advanced OCT retinal image analysis, contributing to… Show more 1.Developed and implemented a complete end-to-end data pipeline in AWS using SageMaker and AWS Batch, including data ingestion, preprocessing, model training, and model registry.2.Combined U-Net with Transformers to create a cohesive system for segmenting and classifying elements in Retinal CT images.3. Deployed pretrained models on AWS Sagemaker, AWS Lambda, Docker, Kubernetes.4.Developed and deployed deep learning models for advanced OCT retinal image analysis, contributing to scientific advancements in the early detection and diagnosis of eye diseases. These models outperformed previous methods in accuracy and efficiency, and provided scalable and accessible solutions for clinicians and researchers globally by being implemented on the cloud infrastructure. Show less

1- Contributed to OpenAI's development of the RLHF model for training GPT-4. Enhanced ChatGPT (Tether project), improving factuality and mathematical capabilities. Debugged and corrected Python code outputs. Applied Prompt Engineering methods to enhance model response quality. 2-Designed and developed enterprise search and conversational agents using Gemini-Pro and Google Vertex AI, providing a low-code/no-code environment for complex tasks3- Developed a specialized coding… Show more 1- Contributed to OpenAI's development of the RLHF model for training GPT-4. Enhanced ChatGPT (Tether project), improving factuality and mathematical capabilities. Debugged and corrected Python code outputs. Applied Prompt Engineering methods to enhance model response quality. 2-Designed and developed enterprise search and conversational agents using Gemini-Pro and Google Vertex AI, providing a low-code/no-code environment for complex tasks3- Developed a specialized coding assistant for the Agora platform using Gemini-Pro, integrating retrieval-augmented generation (RAG) and custom embeddings for efficient document retrieval. Implemented context-aware code generation from proprietary codebases, enhancing developer productivity with tailored AI-driven solutions Show less

Istanbul, Turkey

1-Fine-tuned the Llama 3.1 8B model on a dataset of biomedical question-answer pairs using QLoRA (Quantized Low-Rank Adaptation)Built a biomedical knowledge base by indexing over 50,000 biomedical abstracts using Faiss library for efficient similarity search2-Implemented a retrieve-and-read QA system by first retrieving relevant passages from the Faiss index based on the question, then reading and extracting the answer span from the passages using the fine-tuned LLama… Show more 1-Fine-tuned the Llama 3.1 8B model on a dataset of biomedical question-answer pairs using QLoRA (Quantized Low-Rank Adaptation)Built a biomedical knowledge base by indexing over 50,000 biomedical abstracts using Faiss library for efficient similarity search2-Implemented a retrieve-and-read QA system by first retrieving relevant passages from the Faiss index based on the question, then reading and extracting the answer span from the passages using the fine-tuned LLama model3-Containerized the model and inference pipeline using Docker for easy deployment,Created an AWS Lambda API to serve4-Developed a mobile health application that gives the user personal analysis and recommendations for upper airway infections such as Flu or bacterial infection Show less