Hire Yong L. - Machine Learning Engineer

Generative Recommendation Arch Codesign (GRACE) team in Ads Ranking

Developed an efficient algorithm for Intent Recognition that leverages BERT’s embedding which requires no fine-tuning and a small dataset, producing models that perform with high precision (90%+) with an acceptable recall (45%+) which can be served using the above system. Led the False Negative ML Strategy, ensuring that the attack detection system solves at least 30% of previously reported false negatives each month through the FN Coverage metric construction for observability and the addition of FN-targeted models by using a targeted sampling approach to maximize model sensitivity. Co-designed and spearheaded the adoption of the internal Auto-Retraining Platform as the ML training platform, ensuring ML model reproducibility and ML model recency across the ML engineering team with minimal human intervention, optimizing efficiency in model updates and deployment.