RAG Applications in Industry

The speakers, Deniz Sezin Ayvaz and Chinara Aliyeva from PricewaterhouseCoopers (PwC), will present two different demos under the same topic.

Bio – Deniz Sezin Ayvaz

As a data science consultant, I help organisations develop data-driven strategies, leveraging analytics, AI, and digital transformation to drive impact. My experience spans diverse industries, including retail, financial services, healthcare, and insurance, where I’ve worked on projects such as sales forecasting, marketing budget optimisation, dynamic pricing, and customer behaviour analysis.

I hold a Bachelor’s degree in Statistics and Operations Research and a Master’s in Computer Engineering, combining analytical rigor with technical expertise. To further strengthen my capabilities, I’ve earned certifications such as Machine Learning DevOps Engineer nano degree. My commitment to ethical AI led me to participate in a data study group at the Alan Turing Institute, where I contributed to a published report on mitigating bias in machine learning models for financial data.

Beyond my professional work, I am passionate about continuous learning and fostering inclusivity in tech. Having relocated from Istanbul to London, I understand the challenges of adapting to new environments and building a career abroad. Outside of work, I enjoy tennis and singing.

Abstract – Deniz Sezin Ayvaz

The presentation showcases a Streamlit web app that integrates generative AI (LLMs) to enhance the interpretation and interaction with machine learning model results. Focused on a marketing mix model, the app enables users to convert natural language queries into actionable insights, facilitating a deeper understanding of how marketing channel spend impacts revenue through a time series forecasting model. Key features include an attribution plot using SHAP values to explain model predictions, and an LLM-powered interface that allows users to ask specific questions about the model's output. The app makes advanced analytics more accessible, empowering both technical and non-technical users to derive meaningful insights and make data-driven decisions.

Bio – Chinara Aliyeva

I am a dedicated Data Scientist with over 8 years of experience in AI, advanced analytics, and software engineering. Currently part of PwC’s Generative AI Pod and an active member of Woman in Data, I lead the design and development of innovative solutions with Large Language Models (LLMs) to transform business operations.

My work focuses on building scalable automation frameworks that enhance decision-making and drive operational efficiency. I’m also passionate about fostering collaboration and knowledge-sharing, regularly leading team trainings and organizing AI hackathons to promote continuous learning.

I hold an MSc in Computational Finance from University College London and a BSc (Hons) in Information Technologies and Systems Engineering. My technical expertise is in AI, NLP (LLMs, BERT, RAG), Python, Azure OpenAI, and Databricks.

Abstract – Chinara Aliyeva

Unlocking Insights from Bulky Financial Statements Using RAG, LLMs & Sentiment Intelligence

The presentation is about a bespoke AI-powered tool designed to process and analyze large-scale financial statements, often spanning hundreds of pages. It leverages Retrieval-Augmented Generation (RAG) alongside Gemini and OpenAI models to enable advanced information extraction and context-aware question answering. The system delivers responses grounded in document evidence, enriched with reasoning capabilities to explain the "why" behind each answer.

A key feature of the tool is its sentiment analysis engine, which identifies and explains tone—positive, neutral, or negative—across financial-related themes. This is achieved through a combination of the FinBERT model and a Large Language Model acting as a judge, which refines and validates sentiment outputs to ensure reliability and interpretability.

By combining deep retrieval, language understanding, and sentiment reasoning, the tool transforms complex financial documents into accessible, actionable insights—enhancing decision-making across finance, risk, and sustainability domains.