> Markdown version of [/videos/1088-unleashing-the-power-of-ai-to-prevent-financial-crime](https://www.wearedevelopers.com/videos/1088-unleashing-the-power-of-ai-to-prevent-financial-crime). Every page supports `.md` or `Accept: text/markdown`. Links point to the HTML versions so they work for humans too. Agent guide: [/agents.md](https://www.wearedevelopers.com/agents.md).

---

# Unleashing the power of AI to prevent financial crime

When banking compliance ruled out external APIs, engineers had to get creative. Learn how their dual-layered ML pipeline uses offline NLP to stop 70% of fraud in real time.

- **Speakers:** [Doraly Chezeu Sukem](https://www.wearedevelopers.com/@doraly-chezeu-sukem), [Esha Agrawal](https://www.wearedevelopers.com/@esha-agrawal)
- **Event:** World Congress 2024
- **Published:** August 20, 2024
- **Duration:** 29:40
- **URL:** https://www.wearedevelopers.com/videos/1088-unleashing-the-power-of-ai-to-prevent-financial-crime

## Summary

This presentation explores the practical applications of AI in the banking sector, focusing on both security operations and customer experience. Moving beyond simple compliance, the narrative highlights two core engineering initiatives at ING: preventing financial crime through advanced fraud detection and helping users track personal expenses by intelligently categorizing unstructured transaction data.

To combat credential theft and unauthorized account takeovers, the engineering team deployed a dual-layered transaction monitoring pipeline. First, a rule-based filtering system checks for immediate warning signs, such as new transaction approvers or internal routing red flags. Suspicious events are pushed to an Elasticsearch database, where a machine learning anomaly detection model compares real-time metrics—like IP geolocation and transaction frequency—against historical user behavior. By executing this logic in real time, the pipeline successfully intercepts nearly 70% of fraudulent transfers before funds leave the bank.

Shifting to customer experience, AI is leveraged to categorize non-card transactions based on user-entered text descriptions. The engineering team encountered a significant localization challenge: minority user languages were degrading model performance within a predominantly French historical dataset. Because strict banking compliance regulations and "missing context" risks ruled out external translation APIs, they built an efficient offline inference solution. By using a count vectorizer to isolate high-frequency terms, the team generated a lightweight, keyword-driven translation mapping framework. This scalable NLP approach significantly stabilized machine learning accuracy across multiple languages while maintaining exceptionally low prediction latency.

**Keywords:** financial crime prevention, fraud detection pipelines, transaction monitoring, credential theft, anomaly detection models, rule-based transaction filtering, Elasticsearch, transaction categorization, merchant category code mapping, financial natural language processing, machine learning localization, count vectorizer mapping, offline translation models, predictive customer behavior, banking compliance operations

## Chapters

1. **Speaker introductions and overview of ING bank** (00:02) — An introduction to the speakers, ING's scale in the banking industry, and technical career benefits.
1. **Overview of AI applications in banking** (04:28) — How banks apply artificial intelligence across transaction monitoring, credit risk, and customer experience.
1. **Understanding the credential theft process in banking** (05:20) — The mechanics of how fraud actors use phishing to steal credentials and initiate unauthorized transactions.
1. **Two-step fraud detection with rules and machine learning** (10:08) — Combining rule-based processing with an anomaly detection model to identify fraudulent transactions in real time.
1. **Improving customer experience with transaction categorization** (15:48) — Using artificial intelligence to organize unstructured transaction logs into clear spending categories.
1. **Categorization workflows for card and non-card transactions** (18:33) — Mapping merchant codes for card purchases and applying machine learning to non-card transaction descriptions.
1. **Addressing the challenge of multilingual transaction descriptions** (21:28) — Machine learning performance degrades when classifying transaction strings written in minority languages.
1. **Evaluating modeling solutions for multilingual text classification** (22:49) — Why single composite models, per-language models, and external translation APIs fail in regulated banking environments.
1. **Implementing offline translation mapping for model features** (25:15) — Translating minority language keywords into a majority language base using a pre-trained offline dictionary.

## Related Moments

- [Utilizing artificial intelligence to scale financial asset classification](https://www.wearedevelopers.com/videos/100125-tech-as-a-profit-center-maximizing-p-l-impact-by-research-funding-and-capitalization) (from "Tech as a Profit Center – Maximizing P&L Impact by Research Funding and Capitalization")
- [Securing heterogeneous legacy payment infrastructure against AI](https://www.wearedevelopers.com/videos/100331-fighting-the-next-wave-of-cybercrime) (from "Fighting the Next Wave of Cybercrime")
- [Evaluating unsupervised anomaly detection model performance in banking](https://www.wearedevelopers.com/videos/111-detecting-money-laundering-with-ai) (from "Detecting Money Laundering with AI")
- [Developing AI solutions for large banking scales](https://www.wearedevelopers.com/videos/1627-pioneering-ai-assistants-in-banking) (from "Pioneering AI Assistants in Banking")
- [Navigating technology adoption across business and IT functions](https://www.wearedevelopers.com/videos/1835-using-ai-in-talent-teams-what-works-what-doesn-t) (from "Using AI in Talent Teams: What Works, What Doesn’t")
- [Overcoming limitations of traditional anti-money laundering systems](https://www.wearedevelopers.com/videos/111-detecting-money-laundering-with-ai) (from "Detecting Money Laundering with AI")

## Related Articles

- [Coffee with Developers - Maria Apazoglou - Making AI understandable for all in production](https://www.wearedevelopers.com/magazine/475-coffee-with-developers-maria-apazoglou-making-ai-understandable-for-all-in-production)
- [Stephan Gillich - Bringing AI Everywhere](https://www.wearedevelopers.com/magazine/489-stephan-gillich-bringing-ai-everywhere)
- [MLOps And AI Driven Development](https://www.wearedevelopers.com/magazine/82-mlops-and-ai-driven-development)
- [MLOps – What’s the deal behind it?](https://www.wearedevelopers.com/magazine/125-mlops-what-s-the-deal-behind-it)

## Related Jobs

- [AI & Machine Learning Engineer (all genders)](https://www.wearedevelopers.com/jobs/48217-ai-machine-learning-engineer-all-genders) at **msg**
- [AI Software Engineer (Germany)](https://www.wearedevelopers.com/jobs/48317-ai-software-engineer-germany) at **Sunhat**
- [AI Operations Manager (all genders)](https://www.wearedevelopers.com/jobs/48263-ai-operations-manager-all-genders) at **envelio**
- [Penetration Tester / Red team Specialist](https://www.wearedevelopers.com/jobs/ext/293774-penetration-tester-red-team-specialist) at **Raiffeisen Bank International AG**
- [Security Architect - AI](https://www.wearedevelopers.com/jobs/ext/1581899-security-architect-ai) at **ZEISS Group**
- [Machine Learning Engineer](https://www.wearedevelopers.com/jobs/ext/1355348-machine-learning-engineer) at **TWILIO**