Apply on Employer Site

Coupang Fulfillment Services · 12 hours ago

Senior Machine Learning Engineer, Generative AI Search

Mountain View, USA

Full-time

Onsite

Mid, Senior Level

$159K/yr - $204K/yr

2+ years exp

Coupang is one of the fastest-growing e-commerce companies, known for its innovative approach to commerce. The role involves building advanced search systems and improving machine learning models to enhance customer experiences.

DeliveryLogisticsWarehousing

Responsibilities

Build end-end Search & Discovery systems and help design our latest Gen AI based search engine

Ability to drive key developments and overcome technical blockers

Ensure solid ML design and operational excellence

Be metrics driven and identify big opportunities

Build strong collaborations with partner teams

Qualification

Machine LearningDeep LearningPythonStatistical ModelingJavaEmbedding-based RetrievalA/B TestingFeature EngineeringCollaborationProblem Solving

Required

Bachelor's degree in computer science, electrical engineering, mathematics, statistics or closely related fields

2 years of professional experience in applied machine learning

Experience in machine learning, deep learning, and statistical modeling

Proficiency in Python and/or Java, with experience in building production grade ML systems

Preferred

Retrieval: Uses embedding-based retrieval (EBR) with Approximate Nearest Neighbor (ANN) search (e.g., FAISS, HNSW). Familiar with advanced techniques like contrastive learning or adaptive retrieval

Ranking: Know SOTA models like DCNv2, DeepFM, FFM and Generative Recommendation. Explores multi-objective loss functions (e.g., optimizing engagement, diversity, and revenue)

Negative Sampling: Understands and applies standard in-batch and hard negatives to avoid overfitting to exposed distribution for retrieval and ranking

Model Evaluation: Discusses A/B testing, offline metrics (NDCG, MRR, precision@k), and online impact metrics (CTR, conversion rate)

Cold Start & Exploration: Discusses meta-learning, few-shot learning, or reinforcement learning (multi-arm bandit and usage of randomization experiments)

Scalability: Aware of serving constraints and may mention precomputing embeddings or caching strategies

Feature Engineering: Uses standard user, item, query-item engagement features, and contextual features