Machine Learning Engineer III, Search Relevance

Box•Redwood City, CA

13h•$175,500 - $219,500•Hybrid

About The Position

The Search Relevance team at Box powers discovery across billions of files, enabling customers to find the right content quickly, securely, and intelligently. As we expand into a new era of AI-powered content understanding, we’re investing in the foundation that makes great search possible: reliable systems, strong signals, and models that learn from real-world usage. This is a rare opportunity to work at the intersection of information retrieval science, applied machine learning, and large-scale distributed systems. You’ll be building the infrastructure that powers intelligent content discovery for Fortune 500 companies—where milliseconds matter, relevance is measurable, and your experiments directly impact how millions of users work. We’re looking for a Machine Learning Engineer III to improve search quality end-to-end—signals, ranking, retrieval, and evaluation—while building scalable, low-latency services that serve queries in real time. You’ll collaborate with senior engineers, Product, Data, and Infra partners to productionize modern retrieval techniques and experimentation frameworks that directly impact how millions of users work.

Requirements

3+ years of industry experience building backend or distributed systems, with production ownership of services or data pipelines.
Proficient in at least one of: Java, Scala, C++, or Python; comfortable writing production-grade Python is a plus.
Exposure to search, ranking, recommendations, or applied ML in production; understand the basics of training-to-serving workflows.
Experience with data pipelines, message queues, or streaming systems (e.g., Kafka, Pub/Sub) and near real-time processing.
Familiarity with cloud-native microservices, CI/CD, observability, and performance tuning.
BS in Computer Science or related field, or equivalent practical experience.
Pragmatic, metrics-driven mindset—eager to experiment, measure impact, and iterate quickly in collaboration with partners.

Nice To Haves

Experience with Elasticsearch, Solr, Lucene, or custom search systems; understanding of inverted indexes and scoring functions.
Knowledge of relevance tuning, learning-to-rank concepts, and offline/online experimentation practices.
Exposure to vector search, dense/sparse embeddings, and hybrid retrieval architectures.
Familiarity with IR fundamentals (BM25, TF-IDF, multi-stage retrieval) and query understanding.
Experience with Kubernetes/Terraform and a major cloud (GCP/AWS/Azure).
Practical exposure to PyTorch or TensorFlow; LLM familiarity helpful but not required.

Responsibilities

Design, build, and iterate on components for ranking, retrieval, and recommendations that improve measurable relevance and latency.
Implement production features leveraging embeddings, semantic/hybrid search, and LLM-enabled retrieval under mentorship and design guidance.
Contribute to offline/online evaluation, A/B tests, and relevance tuning using metrics such as NDCG, MRR, and precision@k.
Develop reliable, observable microservices and near real-time indexing pipelines across distributed systems.
Own well-scoped projects from design to rollout, writing clear design docs, tests, and operational runbooks.
Improve data and feature pipelines (batch/streaming) to ensure quality, freshness, and end-to-end performance.
Document patterns and contribute to team best practices that raise the bar on code quality and reliability.
Participate in our on-call rotation, available at all times while on-call to help respond to and triage any issues that arise.