Software Engineer, ML Infrastructure, Content Retrieval Platform, Level 4

Snap
CA, USA / WA, USA / NYC, USA2026-08-31Full time

About the job

We're looking for a Software Engineer to join the Content Retrieval Platform team, part of the Content ML organization. We develop and operates on large scale retrieval system for content recommendation, partner with Applied ML teams for architecture changes and model developments. We utilize Java/Go as the main indexing and serving languages with performant pieces in C++. We are building agentic repo starting from retrieval root and incubating L0.5, generative retrieval, and redesign of inference feature sourcing.

Responsibilities

Design and optimize infrastructure systems for machine learning workloads at scale and drive reliability and efficiency improvements across Snapchat’s ML Infrastructure

Build and enhance feature generation and serving pipelines that power online inferencing and offline training data generation

Develop high-performance inference systems to ensure fast and efficient AI model serving

Build infrastructure to perform scalable ML model training, evaluation, and inference in the cloud

Build comprehensive data management systems for scalable data collection, labeling, processing, and evaluation

Work closely with ML engineers to deploy cutting-edge models into production

Qualifications

Minimum

Bachelor’s degree in a technical field such as computer science or equivalent experience

2+ years of post-Bachelor’s software development experience; or Master’s degree in a technical field + 1+ year of post-grad software development experience; or PhD in a relevant technical field

Experience building large scale production machine learning systems, distributed systems or big data processing

Preferred

Masters/PhD in a technical field such as computer science or equivalent industry experience

Experience working with ML Training platforms or optimizing AI model inference

Familiarity with ML frameworks such as TensorFlow, PyTorch, Caffe2, Spark ML, scikit-learn, or related frameworks