Machine Learning Engineer (Search & Retrieval Systems)

Role overview

You will own and evolve Wizard’s search and retrieval systems that power the AI shopping agent. The role focuses on building hybrid retrieval pipelines, adaptive ranking models, and a feedback loop that continuously improves product quality metrics.

Key missions

• Evolve the hybrid search pipeline, including:
  • lexical retrieval
  • dense/vector retrieval
  • reciprocal rank fusion
  • multi-stage reranking
  • implementation on Elasticsearch
• Build and train adaptive retrieval models, including lightweight query/category/context-aware components that adjust retrieval behavior.
• Design and productionize learning-to-rank, covering:
  • feature engineering
  • training
  • deployment
  • offline evaluation and A/B testing
• Own the offline enrichment pipeline and ensure end-to-end instrumentation and evaluation.
• Integrate query understanding outputs into retrieval and ranking.
• Measure success via improvements in product quality metrics, pipeline adaptability, and integration of behavioral signals.

Requirements

• 5–8+ years shipping production search, retrieval, or ranking systems
• Strong Python skills and solid production engineering fundamentals (clean, typed, structured code)
• Pragmatic ML approach: simplest effective model, rigorous measurement, iterative shipping
• Offline evaluation methodology experience including nDCG, MRR, precision/recall@k and A/B test design/interpretation
• Hands-on experience with Elasticsearch (or similar: Solr, Vespa, OpenSearch), including indexing and query optimization for hybrid retrieval
• Experience with embeddings and vector search (dense retrieval, ANN indexing, embedding fine-tuning)
• Learning-to-rank experience with LightGBM and/or similar (e.g., XGBoost, LambdaMART)

Nice to have

• ONNX for model export/deployment (mentioned in the productionization flow)