Deep Learning Intern (RecSys for Anti-Fraud)

Intern in Sber's anti-fraud team for the development of deep learning-based recommendation system (RecSys) models operating in real-time under high load.

Tasks:

• Adapt SASRec/HSTU/two-tower models for anti-fraud
• Work with transaction chains and client profiles
• Predict on the last transaction

Requirements:

• Proficient Python, PyTorch (built models from scratch)
• Knowledge of architectures: Transformer, YouTubeDNN, SASRec, Bert4Rec, HSTU, ARGUS (at least a couple)
• Ability to work with tabular features for DL
• Understanding of Negative sampling and loss functions for ranking/classification

Conditions:

• Paid internship 3–6 months
• Salary 81 thousand rubles/month
• Employment from 20 hours/week
• SberCity office

Skills: Python / PyTorch / Transformer / YouTubeDNN / SASRec / Bert4Rec / HSTU / ARGUS / RecSys / BigData