The success of GenAI apps is decided by the accuracy of their responses. Using Retrieval Augmented Generation (RAG), you can improve accuracy by grounding GenAI app responses in your data. In this session, explore advanced RAG techniques in Azure Database for PostgreSQL including new vector search algorithms, parameter tuning, hybrid search, semantic ranking, and the GraphRAG approach. See how customers are using these techniques to deploy corporate development platform for GenAI apps.

Maxim Lukiyanov
Principal PM Manager
Maxim is leading Generative AI & Developer Experience PM team in Postgres on Azure service. He has extensive experience in building Data and AI cloud services and has been with Microsoft for 15 years.

Orhun Oezbek
Software Engineer
Orhun Özbek is the GenAI Lead and a software engineer at UBS, specializing in building reusable and scalable GenAI platforms and providing business solutions for banking AI use cases. He leads the UBS RiskLab Model Services GenAI team which develops the RiskLab AI Common Ecosystem platform which streamlines AI development and deployment across UBS through innovative RAG, AI Agents, Benchmarking and Guardrail functionalities.
Orhun will present RiskLab VEGA (Vector Embedding Governance Application) which is a cornerstone of the RiskLab AICE platform. RiskLab VEGA leverages Azure PostgreSQL PGVector to offer governed and self-service Vector Store RAG development and benchmarking capabilities for AI developers. His presentation will include a deep-dive into the challenges of processing complex financial documents and explore advanced indexing and retrieval techniques.
With a background as an ML/AI data scientist, Orhun now utilizes his expertise to build and optimize AI/ML platforms. He is passionate about creating reusable IT systems and advancing AI and machine learning capabilities in the financial sector.
Orhun will present RiskLab VEGA (Vector Embedding Governance Application) which is a cornerstone of the RiskLab AICE platform. RiskLab VEGA leverages Azure PostgreSQL PGVector to offer governed and self-service Vector Store RAG development and benchmarking capabilities for AI developers. His presentation will include a deep-dive into the challenges of processing complex financial documents and explore advanced indexing and retrieval techniques.
With a background as an ML/AI data scientist, Orhun now utilizes his expertise to build and optimize AI/ML platforms. He is passionate about creating reusable IT systems and advancing AI and machine learning capabilities in the financial sector.

Jay Yang
Executive Director, RiskLab
Jay Yang is an Executive Director and Distinguished Engineer working for UBS. He leads the UBS RiskLab Model Service team which manages the largest data science lab within UBS serving more than 1,200 data scientists and ML/AI developers.
UBS RiskLab builds a suite of ML/AI and Data Analytic common platforms and solutions leveraging Azure native services to enable self-service capabilities for users to expedite Model/AI development and streamline production deployment. The RiskLab AI Common Ecosystem (AICE) platform, an integral platform of the UBS AI Playbook, provides a governed and tenanted RAG environment, streamlined AI Agent standardization, registration and deployment solutions, and end-to-end ML model continuous integration, training, deployment and monitoring processes. In addition, Risklab data science platform offers governed corporate data mesh access and containerized big data compute environments.
Jay has over 20 years experiencing working for technology startups, fintech and financial institutions. He has a passion for building reuseable IT systems and ML/AI platforms.
UBS RiskLab builds a suite of ML/AI and Data Analytic common platforms and solutions leveraging Azure native services to enable self-service capabilities for users to expedite Model/AI development and streamline production deployment. The RiskLab AI Common Ecosystem (AICE) platform, an integral platform of the UBS AI Playbook, provides a governed and tenanted RAG environment, streamlined AI Agent standardization, registration and deployment solutions, and end-to-end ML model continuous integration, training, deployment and monitoring processes. In addition, Risklab data science platform offers governed corporate data mesh access and containerized big data compute environments.
Jay has over 20 years experiencing working for technology startups, fintech and financial institutions. He has a passion for building reuseable IT systems and ML/AI platforms.