Sinan Kefeli is an AI scientist with a focus on natural language processing (NLP), machine learning, and computer vision. At Verizon, he develops models to analyze customer call transcripts, improve customer interactions, and answer complex queries using large language models (LLMs). His work includes fine-tuning open source models like Mistral and Flan-T5, building graph databases, and creating tools for real-time information retrieval.

Prior to Verizon, Sinan earned his PhD in Physics from Caltech, where he applied advanced machine learning techniques to large-scale scientific data pipelines.

Sinan uses tools like Python, PyTorch, Neo4j, and Docker to build reliable and efficient AI systems. He focuses on practical solutions that are clear, measurable, and effective.

 

Presentations

22x

Enhancing Occupational Classification and Seniority Prediction from Job Descriptions Using Open-Source Tools and Embedding Techniques

Accurate job classification is essential for understanding labor market trends and informing policy decisions, but modern job descriptions are complex and varied. We explore how open-source AI tools like PyTorch and HuggingFace models can improve mappings of Standard Occupational Classification to job descriptions. Using these technologies, we develop methods to better interpret and classify unstructured job data, leading to more precise occupational analysis. Join us to learn how open-source AI can enhance workforce analytics and how you can apply these tools in your own projects.

See Presentation