Open Source in Finance Forum London 2025

Effective trade and market surveillance represents a cornerstone of financial market integrity, providing the foundation of trust upon which the entire financial system depends.
Given huge and ever-growing volumes of heterogeneous time-series trade and market data coming from different sources in real-time, being able to capture and effectively process in-memory, real-time, and historical time-series data has become a key component for any effective surveillance solution.
We plan to present and demonstrate how Timebase was leveraged in the following case studies
• Mission critical trade surveillance solution that addresses bank’s trading and regulatory compliance
• Demo that examines a market manipulation event to showcase Timebase out of the box capabilities in solving market surveillance challenges

Meet OS-Climate’s PhysRisk library, an open, transparent, and auditable framework that integrates high-quality climate data with advanced risk modeling techniques. Participants will explore methodologies for evaluating physical climate risks under different climate scenarios across diverse sectors, learning how to integrate geospatial data, vulnerability models, and risk-scoring techniques into their risk assessments.

Attendees will gain actionable insights into climate risk quantification, regulatory compliance, and strategic risk management—leveraging open-source tools to build climate resilience across industries.

A practical banking sector use case will illustrate how financial institutions assess climate risks at two key levels: evaluating how climate hazards impact corporate counterparties, integrating risk models to assess financial exposure and credit risk; and analyzing the resilience of physical assets linked to financial agreements, using location-based vulnerability modeling to quantify potential damages.

IT professionals attending who are new to the topic will be equipped to take sufficient knowledge back to their respective internal climate teams.

The EU Artificial Intelligence Act represents a major regulatory milestone, embedding environmental sustainability into AI governance. It introduces principles for developing AI systems sustainably, mandates energy consumption logging for high-risk AI, and requires lifecycle environmental impact assessments. Key provisions emphasize green AI development, mandate tracking energy and resource use and enforce efficiency standards for foundational models. However, despite these advancements, the Act's final version leans on voluntary standards, limiting enforceability. It largely focuses on energy consumption, overlooking broader ecological concerns like water use, mineral extraction, and electronic waste. While gaps remain, the Act is a step toward sustainable AI, reinforcing the need for responsible innovation. Future efforts must strengthen enforcement to ensure AI development aligns with global environmental goals.

Financial institutions require solutions for creating and sharing high quality data sets, internally and externally for analytics and AI/ML.

In this presentation we will introduce one of the foundational pieces of OS-Climate Data Commons: its data platform that provides its constituents - producers and consumers of data - with a toolbox for sharing datasets, as well as easily creating and sharing new datasets on the platform.

Bringing concepts such as Data-as-Code, GitOps, and CI/CD to data ingestion, transformation and publishing, and based on the conceptual framework of the Data Mesh, we will present and demonstrate a Data Platform built with open source components such as Apache Airflow, Spark, Apache Iceberg, Trino, Unity and more and running cloud-natively on Kubernetes. This provides solutions for a wide range of problems faced by services institutions in the data space - from self-servicing data infrastructure for running ETL pipelines, to decentralizing data product ownership while preserving security and regulatory compliance in a federated governance model.

Come see it at work in a practical exercise!

TRAC DAP is an incubating FINOS project that will publish v1.0 around conference time.

Thesis:
Risk and Finance functions are poorly served by established data science patterns because most of their analytics are ‘structural’, meaning they involve applying complex semi-rigid operations to generate outputs with specific, often regulated, definitions.

Examples of structural analytics in banking include regulatory capital, accounting models, stress testing, credit scoring, liquidity reporting, ALM (a similar list for insurance)

Version control, lineage, repeatability, auditability and the ability to experiment within a defined structure are key capabilities. These are not the capabilities modern data science solution prioritise.

Traditionally, banks built/bought a calculation engine per use case, with aspects of the model or data structure encoded in it.
TRAC D.A.P. takes a different approach. It offers a universal model orchestration framework built around a common metadata model, that works for all structural analytics.

Session objectives:
- Define structural analytics
- Explain how we generalised a solution
- Explain how people can leverage and extend TRAC D.A.P.

How we leveraged AI to enhance ESG analysis while maintaining human judgment. InferESG, an open-source tool developed with abrdn using FinTech Scotland research grant funding, demonstrates a practical approach to this challenge. Using an AI workflow and multi-agent AI architecture, the system augments analyst capabilities in assessing materiality and detecting potential greenwashing.
We'll cover:

How AI can assist in parsing complex ESG reports and frameworks (Corporate Sustainability Reports, ESG Ratings, TNFD, GRI)
Practical approaches to materiality assessment using AI
Techniques for detecting subtle forms of greenwashing
Real-world validation results from testing with major corporations
Balancing AI automation with human expertise
Open source architecture for community enhancement

The session includes demonstrations of the greenwashing detection and materiality assessment, showing how analysts can interact with the system to investigate ESG claims and verify corporate sustainability statements.

Mainframe computers are old, mostly depicted with black and white photos of operators loading tape drives or else history boards telling of their use in the moon landings. Despite their age however, they remain the backbone of financial transactional IT, and are undergoing a revival powered by by a large stack of open source languages and packages. This talk will cover how the Linux Foundation Open Mainframe Project is at the forefront of this, bringing test tools like Galasa, Visual Studio Code and IntelliJ plugins from the Zowe project, ports of 300 tools like bash, curl, openssl to give a current linux environment. With languages like python, Java, node.js being on the platform, as well as the LinuxOn Z environment this is no longer your grandparent's machine. We'll show how open source is at the driving edge of bringing the platform into the 21st century, and also how the blend of open source and enterprise computing cope when they meet.