Senior Data Scientist - LLMs for Fraud Detection

Job Description

Job Description

About DataVisor

DataVisor is the world's leading AI-powered Fraud and Risk Platform that delivers the best overall detection coverage in the industry. With an open SaaS platform that supports easy consolidation and enrichment of any data, DataVisor's fraud and anti-money laundering (AML) solutions scale infinitely and enable organizations to act on fast-evolving fraud and money laundering activities in real time. Its patented unsupervised machine learning technology, advanced device intelligence, powerful decision engine, and investigation tools work together to provide significant performance lift from day one. DataVisor's platform is architected to support multiple use cases across different business units flexibly, dramatically lowering total cost of ownership compared to legacy point solutions. DataVisor is recognized as an industry leader and has been adopted by many Fortune 500 companies across the globe.

Our award-winning software platform is powered by a team of world-class experts in big data, machine learning, security, and scalable infrastructure. Our culture is open, positive, collaborative, and results-driven. Come join us!

Why this role
Fraudsters now use generative AI, and we’re using it back. At DataVisor you’ll build and fine-tune large models that run on billions of real-world events and catch attacks traditional rules miss. You’ll take models from idea to production and see the results in live fraud stopped for some of the world’s largest banks, fintechs and platforms. If you want to apply cutting-edge LLM work to a hard adversarial problem with real stakes, this is that job.

What you’ll do

Build and fine-tune large models

• Fine-tune and adapt LLMs and other foundation models for fraud tasks such as behavior understanding, entity and identity reasoning, and case summarization.
• Design and train large models on sequences, graphs and other multi-modal event data covering users, devices, networks and transactions.
• Build evaluation frameworks for model quality, robustness and adversarial drift, and own model serving at production latency and cost.

Build agentic AI for fraud

• Build AI agents and reusable tools that automate investigation, feature generation and reporting on our proprietary AI platform.
• Turn expert investigator workflows into AI systems that scale across hundreds of customers.

Run ML end to end at scale

• Own the full lifecycle of detection models: exploring data, prototyping, shipping to production and monitoring.
• Engineer features from huge, noisy, imbalanced data (billions of events) using Spark and SQL.
• Combine classic ML (gradient boosting, graph methods) with deep learning and LLMs, and know when each works best.

Investigate real attacks

• Dig into complex fraud cases, reconstruct how attackers operated, and feed the findings back into models and products.
• Write clear technical reports and trend analyses for customers, executives and engineering teams.

Requirements

• Master’s or PhD in Computer Science, Statistics, Mathematics or a related quantitative field.
• 3+ years of hands-on experience in at least one of:
• LLMs and large models: fine-tuning, post-training, or building and training large models in production or serious research, or
• Fraud and adversarial risk: fraud detection, cybersecurity, trust & safety, payments or marketplace risk.
• Strong computer science fundamentals: data structures, algorithms and system design, plus software engineering habits like clean, tested, maintainable production code in Python. You’re as comfortable writing efficient code as training models.
• Strong deep learning skills (PyTorch preferred) and a solid grounding in classic ML.
• Production ML experience: you’ve shipped models that real users depend on.
• Strong Python and SQL; experience with Spark or PySpark and with cloud platforms (AWS, GCP or Azure).
• An investigator’s mindset: forming hypotheses, separating signal from noise, and reasoning about adversaries.
• Clear communication with technical and non-technical audiences, including customers and executives.

Nice to have

• Publications at ML, NLP or security venues (NeurIPS, ICML, ICLR, ACL, KDD, USENIX Security and similar).
• Experience with distributed training, inference optimization or retrieval-augmented systems.
• Graph ML or sequence modeling on behavioral data.

Benefits

• Base salary range: $140,000–$170,000, commensurate with experience.
• PTO, Stock Options, Health Benefits