Credit Decision Science
Turn risk signals into lending decisions.
Application score, behavioral score, PD bands, cutoff, limits, pricing, and portfolio monitoring built around real risk-return trade-offs.
Tran Quoc Viet / Charlie · HCMC
Data Scientist focused on credit decisioning, GenAI agents, and ML platforms: turning messy banking data into scoring, automation, monitoring, and production decision systems.
Expertise
Credit Decision Science
Application score, behavioral score, PD bands, cutoff, limits, pricing, and portfolio monitoring built around real risk-return trade-offs.
GenAI & Agentic Systems
Hybrid RAG, multi-agent routing, customer support automation, analytics agents, evaluation, tracing, and human validation loops.
Data & ML Platform
Airflow, dbt, BigQuery, streaming, feature store patterns, production scoring, and AutoML workflows that teams can operate.
What I optimize
Credit scoring workflows from label preparation to OOT validation, calibration, and production scoring
Decision simulations across approval, cutoff, limit, pricing, expected loss, and portfolio guardrails
GenAI agents for customer support, case triage, analytics, retrieval, evaluation, and tracing
Data and ML platform foundations with Airflow, dbt, BigQuery, Kafka, Doris, and feature stores
Latest writing
Collection score không chỉ tìm người dễ trả. Nó là bài toán phân bổ capacity: gọi ai, dùng treatment nào, kỳ vọng thu được bao nhiêu và chi phí có đáng không.
Hai khách hàng cùng PD có thể tạo expected loss rất khác nhau. Khi model score đi vào hạn mức và pricing, DS phải nối probability với exposure, utilization, margin và adverse selection.
Cutoff là một business decision được ngụy trang dưới dạng model threshold. Bài này đọc approval, marginal group, expected loss, profit và rollout guardrail như một bài toán decisioning, không chỉ là chọn ngưỡng.