Florina Regius 플러리나

As Forward Deployed AI Lead on BMO's Commercial Banking GenAI Foundation Program, I architect and deliver enterprise-grade Generative AI solutions in close collaboration with EY, building on AWS Bedrock and Claude to power Retrieval-Augmented Generation (RAG) pipelines and AI Orchestration frameworks. My work spans the full GenAI lifecycle — from platform architecture and developer ecosystem integration to production deployment inside a highly regulated banking environment. In parallel, I lead AI-driven QA Automation initiatives within BMO's B2B Banking & Credit Technology organization, owning use cases in an AI SDLC prompt library, including automation stability and flakiness detection frameworks that reduce testing overhead and improve release confidence. My work bridges GenAI strategy, agentic and RAG-based system design, and hands-on delivery — translating frontier AI capabilities into measurable, production-ready outcomes for one of Canada's largest financial institutions.

Currently leading the development of a Retrieval-Augmented Generation (RAG)-based chatbot for JP Morgan Chase, leveraging LangChain to enhance conversational AI with real-time knowledge retrieval. 🔹 Architecting scalable RAG solutions that ensure accurate, context-aware responses. 🔹 Leading the team in designing and optimizing AI-driven chatbots for financial applications. 🔹 Driving Gen AI innovation within the Center of Excellence (COE), exploring advanced AI applications for enterprise solutions.

I successfully spearheaded the development of a Python automation solution to streamline the processing of over 200+ gene extraction files. Traditionally, the unit testing for 209+ files was conducted manually, a time-consuming and labor-intensive process. Recognizing the inefficiency, I took the initiative to develop a Python-based automation script that extracted and processed data from various JSON files. As a result, I not only completed the unit testing for 250+ files but also delivered it ahead of schedule. This innovation averted potential delays, saving the project from missing its deadline. My automation solution reduced the time required for testing by 10 days, significantly minimizing resource allocation and eliminating the need for additional manpower. Initially, there were reservations about the reliability of the automation. However, my code operated flawlessly, surpassing expectations, and seamlessly handled extra work beyond the initial scope. Despite the challenges and long hours spent developing the automation, the outcome greatly contributed to the project’s success, proving both its robustness and efficiency. Product Development- In house Finance dashboard Developing automation pipelines using Python to extract tables from the necessary portals for developing the dashboard. Also, performing complex SQL queries in Data grip for getting needed columns to display in Power Bi.

Product development - Chatbot : In-house product built using machine learning algorithms with the best accuracy levels. Data ingestion is performed through conversions happening in server.

Facebook Building score-card dashboard in Power Bi where data flows from AWS S3. This dashboard allows the user to know employee performance. Once the data is cleaned and modelled in R-studio. It is uploaded to S3 later. Alert is set in Power bi for data refresh timely. Tiktok Building RCA (root cause analysis) dashboard in Power bi where data flows from AWS S3. This dashboard allows the user to know employee performance. Once the data is cleaned and modelled in R-studio. It is uploaded to S3 later. Alert is set in Power bi for data refresh option. Kronos Building Quicksight dashboard in AWS where Kronos data is used for understanding the anomalies in the behavior towards time keeping. And how it affects their daily performance. Kronos data is transferred to S3 bucket from where data modelling file is triggered using EC2 instances. Followed by postgres sql for data aggregation and anomaly detection sends an email. Along with it, Quicksight dashboard gets refreshed timely as alerts are set. Qorvo Building dashboards in Power Bi where data flows from AZURE. This dashboard allows the user to know the pricing strategy behind every product. As Qorvo is semiconductor company, the pricing varies for various B2B options including Intel, Apple etc. EDA process takes place in python inside data bricks from where Powerbi receives the instant data for refreshing the dashboard periodically.

AWS - Lambda , ECS, ECR, ELB

- Image & Video Analytics - NLP based chatbots - Deep learning based chatbots - Mobile app tracking using Google suite and mixed panel - Power Bi and Tableau Dashboards

Machine learning engineer - Big Data -Classification algorithms -Anomaly Algorithm for data mining - Feeding of data for machine learning models such as GEBI - Missing Alarm analysis

Project-Based Learning Cases Project 1: Social Media Using NLP and Machine Learning, build a model to identify inappropriate tweets that should be removed from a popular social channel’s platform to mitigate social hate and negativity. Project 2: Electronic Commerce The data set provided contains movie reviews given by a video platform’s customers. Perform data analysis on the customers’ movie reviews data set and build a machine learning recommendation algorithm which provides ratings for each of the users. Project 3: Automobile Manufacturing An automotive manufacturer wants the time on its test bench to reduce the time it takes a car to get to the market. Build and optimize the Machine Learning algorithm to solve this problem. Project 4: EdTech Help Simplilearn assess the quality of e-learning videos freely available on YouTube to prepare high-quality and engaging video content for students.