Senior Data Scientist - Detection & Modeling for Finance

TWG AI; New York (NY); On-site; USD 190,000 - 290,000; Full time

Job summary

TWG AI is seeking a Senior Data Scientist for Detection & Modeling to own features and models powering surveillance systems across US exchange and DeFi contexts. You’ll influence architecture as data labels grow and work within a unified modeling practice spanning two detection stacks.

You will design scalable features for time-series and on-chain data, advance anomaly detection, and calibrate scores for cross-venue comparability.

Qualifications

• Strong feature engineering on transactional/time-series data.
• Hands-on anomaly detection and unsupervised methods experience.
• Solid statistical foundations including model calibration.
• Production ML experience with deployment and monitoring.
• Ability to scale modeling across detectors, venues, and markets.
• Experience in fraud, risk, or market-surveillance domains is a plus.
• Knowledge of market microstructure is a plus; on-chain data familiarity a plus.

Responsibilities

• Design and refine features on large-scale financial time-series and on-chain data.
• Develop and improve anomaly-detection models using supervised and unsupervised methods.
• Evolve model architectures as labels accumulate across multiple data stacks.
• Improve detector calibration for trustworthy scores across venue types.
• Collaborate with validation and analytics teams on labeling pipelines.

Skills

Feature engineering

Anomaly detection

Statistical foundations

Production ML

Model scaling

Fraud/risk domain

Market microstructure fluency

On-chain data familiarity

Job description

TWG AI is seeking a Senior Data Scientist for Detection & Modeling to own features and models powering surveillance systems across US exchange and DeFi contexts. You’ll influence architecture as data labels grow and work within a unified modeling practice spanning two detection stacks.

You will design scalable features for time-series and on-chain data, advance anomaly detection, and calibrate scores for cross-venue comparability.