Own the account-abuse scoring system end to end, including detection signals, thresholds, monitoring and recalibration. Build early detection that balances abuse prevention with the cost of false positives and distinguishes fraud from legitimate heavy usage.
Responsibilities:
- Detect abuse rings using payment, IP/ASN, email, registration and behavioural signals, including when reliable device fingerprints are unavailable.
- Design graduated, reversible enforcement actions and define false-positive budgets and appeal paths.
- Build pre-grant abuse gates and partner with Payments, Support and Legal on enforcement policy.
- Measure financial impact, run holdouts where safe and legal, and monitor adversarial adaptation and data-quality issues.
- Provide interpretable scores and evidence suitable for analysts, support agents, auditors and external enforcement.
- Work with raw event, payment and generation data using SQL, Python and BigQuery.
Requirements:
- Experience building abuse, fraud or trust-and-safety detection systems that took action on real users.
- Strong SQL and Python skills, including joins, window functions and cohort replay; experience with gradient boosting and sound judgment about rules versus models.
- Understanding of adversarial modelling, graph/linkage methods, precision-first evaluation and cost-weighted errors.
- Unit-economics literacy and the judgment to escalate when evidence is insufficient.
- Clear written and spoken English, B2+.
Conditions and benefits: Full-time, on-site role in Almaty, Kazakhstan. Relocation support includes a flight and temporary housing. Competitive salary in USD, equity through the stock option program, high autonomy and fast career growth. The hiring process includes a paid one-month on-site trial.