Risk Operations AI Specialist

Responsibilities:

Core Review Duties

Review cases escalated by ROCA where confidence score falls below the auto-decision threshold, using AI-assisted signals within the unified case review UI (no multi-system switching)

Make accurate pass / block / escalate decisions on a daily queue spanning KYC Audit, ATO investigation, Reset 2FA, POA Review, and Withdrawal anomaly case types etc.

Escalate Key User cases (Priority ≥ 75) and high-risk patterns to Senior Agent / Team Lead with clear rationale documentation

Maintain SLA compliance (<4h manual resolution time) while sustaining decision accuracy ≥ 98%

AI Feedback & Pipeline Contribution

Label every reviewed case as TP / TN / FP / FN with structured root-cause annotations — this data directly feeds ROCA's continuous learning pipeline

Identify recurring FP/FN patterns and route them through the correct feedback path:

Type A — Feature weight bias → trigger model retraining / fine-tune

Type B — Threshold miscalibration → update Decision Hub rules (with A/B validation)

 Type C — SOP gap or logic error → update SOP (version-tracked)

Actively contribute to improving Precision (>95%) and Recall (>99%) targets by surfacing systematic signal failures to Data and Model teams

Participate in regular AI model calibration sessions: review misclassification batches, validate ground truth, challenge decision thresholds with data

SOP & Knowledge Management

Maintain and update Confluence SOP pages for assigned case types; SOP coverage rate target ≥ 80% (directly impacts ROCA first-pass accuracy)

Flag SOP coverage gaps when ROCA's reasoning chain reveals unhandled scenarios; draft SOP additions for Team Lead review

Onboard and calibrate BPO agents on AI-assisted review workflows; monitor BPO quality against automated QA benchmarks

Requirements:

1. 2+ years in fraud review, risk operations, or KYC/KYB case management
2. Hands-on experience with ATO (Account Takeover) detection and investigation
3. Comfort working in AI-augmented workflows: you read model outputs, interpret confidence scores, and make decisions based on AI-generated signals rather than raw data alone
4. Structured, precise written communication: you can document a root-cause annotation in <3 sentences that an ML engineer can act on
5. Bilingual English/Mandarin is required to be able to coordinate with overseas partners and stakeholders.
6. An operator's mindset toward AI tools: you leverage AI signals to make faster, better decisions — you master the tool rather than compete with it

Nice to have:

Experience in crypto / Web3 / VASP risk environment; familiarity with blockchain transaction patterns (Sybil attacks, wash trading, mule account indicators)
Exposure to ML model evaluation concepts: Precision, Recall, F1, confusion matrix — you don't need to build models, but you need to speak the language
Familiarity with prompt engineering or LLM evaluation workflows; experience reviewing or testing AI-generated decisions at scale
CAMS, CFE, or equivalent AML / fraud certification
Prior involvement in SOP authoring, QA calibration, or BPO oversight for a risk operations team