Gozde Ozalpay

<p>Initiated and executed a comprehensive AI roadmap to address multi-layer platform fraud, shifting the organization from reactive rules to a proactive ML framework.</p><p> Orchestrated end-to-end alignment across Product, Web, App, and Engineering teams, presenting strategic risk-benefit analyses directly to the CEO's office, fraud, and Legal teams prior to high-impact deployments.</p><p> Engineered an end-to-end detection architecture that slashed manual seller evaluation time from 30 minutes to under two minutes, scaling daily audit throughput by 10x.</p><p> Identified 2 million fraudulent users and purged nearly 500 million fraudulent favorite events (5% of total platform volume) while maintaining near-zero false-positive merchant complaints to ensure operational compliance.</p><p> Reduced fraudulent content visibility by 70% on search pages and 58% on top-category rankings, resulting in a 0.2% absolute CTR improvement for legitimate content.</p><p> Drove algorithmic optimizations across 30 million products, achieving a 0.3% absolute increase in Conversion Rate (CR) that translates directly to massive incremental Gross Merchandise Volume (GMV).</p><p> Architected the foundational consumer risk score model for Trendyol's new fintech lines, shaping downstream strategy for user credit limits and achieving successful integration into live credit screening loops by a major Turkish bank.</p><p> Owned the end-to-end merchant credit-scoring framework, embedding multi-horizon projections directly into partner bank decision loops, which validated a 50% reduction in Non-Performing Loans (NPLs) for approved credits.</p><p> Transformed the seller revenue forecasting pipeline from time-series to regression, achieving a 9% performance increase to reliably evaluate seller financial health.</p><p> Designed autonomous multi-agent loops and MCP servers to automate SQL-to-YAML pipelines, alongside deploying monitoring dashboards for model stability and explainability.</p>

<p>Deployed a high-throughput, real-time fraud detection engine processing 300,000 daily transactions within a 5 million+ aggregate portfolio context.</p><p> Achieved an 80% recall rate within a 1% baseline fraud environment, isolating and blocking 2,400 fraudulent activities daily (876,000+ annually) to protect net revenue and reduce chargebacks.</p><p> Prevented systemic merchant attrition across 100+ key sellers within a single month by delivering actionable retention recommendations to customer experience teams.</p><p> Preserved 2.3 million TRY in transactional revenue and saved over 1.5 million TRY in Customer Acquisition Costs (CAC) by eliminating the need to replace premium merchant tiers.</p>

<p>Joined as the sole data scientist, leading ML models from A to Z for fraud detection, churn prediction, anomaly detection, and customer segmentation.</p><p> Led infrastructure scaling and traffic predictions for Black Friday, managing surges of 1,000 transactions per second with an average ticket size of 1,600 TRY.</p><p> Guaranteed 99% uptime accuracy under extreme load by predicting the maximum load expected; mitigating a 1-minute outage risk directly safeguarded approximately 100 million TRY in transaction volume.</p><p> Monitored end-to-end model performance, executing stability and drift analyses to ensure continuous accuracy across production environments.</p>

<p>Developed PD, EAD, and LGD models for top Turkish and European banks using logistic regression, random forests, and survival models.</p><p> Executed prepayment risk and loss forecasting models with interpretable output to support strict regulatory reporting requirements.</p><p> Processed raw banking data to generate predictive features for customer behavior analysis and enterprise credit scoring.</p><p> Produced actionable dashboards and data reports for executive stakeholders using Tableau and SQL.</p>

<p>Conducted market research and feasibility analyses for large-scale, enterprise real estate investments.</p><p> Prepared macroeconomic correlation studies and Highest & Best Use (HBU) reports for institutional investors.</p><p> Created comprehensive financial models, including NPV, cash flow forecasting, and sensitivity analyses.</p>