VII Course on Machine Learning and Central Banking (practical)

The in-person phase of the VII Course on Machine Learning and Central Banking was held in Panama City from September 2 to 4, 2026. The event brought together 25 participants from 15 institutions across 14 countries, who worked on 18 use cases linked to the core functions of central banks and supervisory authorities.

Rather than following a conventional lecture-based course format, this phase was designed as an intensive working workshop. Participants began with previously identified institutional challenges and moved forward through brief presentations, guided discussion, programming, peer collaboration, feedback sessions, and support from the facilitators.

The cases covered macroeconomic forecasting, inflation and news analysis, foreign exchange markets, financial stability, financial inclusion, payments, fraud, operational risk, customer analytics, automated classification, and multi-agent architectures, among other topics. This diversity encouraged the exchange of methods and experiences among institutions at different stages of analytical maturity.

As a result, the teams refined their working questions, reviewed the availability and relevance of their data, discussed methodological approaches, advanced their prototyping workflows, and received feedback to guide the next stages. The interest expressed by participants creates an opportunity for CEMLA to follow up on the projects and foster a regional community of practice.

• Name: VII Course on Machine Learning and Central Banking — in-person phase
• Dates: September 2, 3, and 4, 2026
• Location: Panama City, Panama
• Format: Applied in-person workshop structured around an agile use-case methodology
• Moderator: Gerardo Hernández del Valle, CEMLA
• Facilitators: Gabriela Alves Werb, Timur Sattarov and Sebastian Seltmann, Deutsche Bundesbank
• Participation: 25 representatives from 15 institutions and 14 countries

The purpose of the in-person phase was to help participants transform a specific institutional need into a better-defined and technically viable machine learning use case. The emphasis was on practical application: each team worked with its own problem, available data, and operational constraints.

The methodology combined four elements: initial presentation and comparison of the cases; successive sprints for ideation, programming, and prototype workflow development; rotating feedback among peers and facilitators; and final presentations to communicate progress, methodological decisions, and next steps.

September 2: Definition and first sprint

• Welcome and introduction to the workshop methodology.
• Presentation of the participants’ use cases and guided group discussion.
• Sprint 1: ideation and development of the initial prototype workflow.

September 3: Programming, collaboration, and feedback

• Sprint 2: programming session and peer collaboration.
• Feedback Coffee: rotating exchange of observations and recommendations.
• Sprint 3: continued programming and refinement of the projects.

September 4: Coaching and final presentations

• Open coaching session and rotating feedback.
• Brief participant presentations on progress and next steps.
• Workshop closing and recap.

In addition to the technical sessions, the program included three activities that supported group integration and professional exchange in a more informal setting:

1. Opening reception. A welcome gathering and first opportunity for participants, facilitators, and organizing teams to connect.
2. Visit to the Museo del Banco Nacional de Panamá. A cultural and institutional activity that offered insight into Panama’s banking and financial history.
3. Closing reception. A final gathering to recognize the work completed, strengthen ties, and bring the in-person experience to a close.

Problem definition. Teams refined the objective, scope, and institutional value of their use cases.

Technical progress. Participants discussed data, variables, methods, and prototyping workflows, with dedicated programming sessions.

Peer learning. Group rotation exposed each project to perspectives from other institutions and functional areas.

Institutional applicability. The projects addressed real policy, operational, supervisory, and analytical needs.

Regional network. The technical exchanges and complementary activities strengthened professional ties among participants.

The experience was highly positive. Combining real cases, intensive work, and specialized support encouraged active participation and highlighted projects that were notable both for their thematic diversity and their potential for implementation within the participating institutions.