Senior ML Operations Engineer II - Ads Platform

Fetch•US - Remote,

About The Position

Fetch is building AI systems for personalization, relevance, and ranking across its Ads Platform. The Senior MLOps Engineer owns the operational foundation for moving ranking and recommendation models from experimentation into reliable production systems. The role partners with Data Science, Product, Platform, and Engineering across the model lifecycle, including reproducible training, validation, deployment, versioning, monitoring, retraining, and retirement. This hands-on technical leadership role sits at the intersection of ML infrastructure, personalization, and large-scale distributed systems. The engineer builds reusable pipelines, improves observability and real-time learning workflows, establishes architecture and reliability standards, and makes model health and progress accessible through a shared model hub. The work supports a more adaptive Ads Platform that responds to changing user behavior.

Requirements

5+ years of experience building and operating production ML systems beyond initial deployment, including monitoring, retraining, incident response, and continuous improvement.
Strong production software development skills in Python and proficiency with SQL.
Hands-on Spark/PySpark experience with large-scale distributed data processing, including feature engineering, historical aggregations, incremental processing, and training dataset construction.
Experience with modern data lake or lakehouse architectures, including S3, Apache Iceberg, or equivalent technologies.
Experience with workflow orchestration for scheduling, dependency management, retries, backfills, and recovery.
Hands-on experience building infrastructure with Terraform.
Experience deploying and operating containerized production workloads on Kubernetes.
Experience building reusable training, evaluation, or deployment infrastructure used by multiple models, scientists, or teams.
Experience implementing model and dataset versioning, experiment tracking, lineage, and reproducible releases.
Ability to distinguish service or infrastructure failures from changes in model effectiveness.
Experience building CI/CD pipelines, monitoring, alerting, rollback mechanisms, and recovery procedures for production ML systems.
Demonstrated ability to balance model quality, delivery speed, reliability, and infrastructure cost.
Strong cross-functional communication and experience partnering with data scientists to preserve scientific validity as models move into production.
Ability to explain a production incident or model degradation issue, how it was detected, and what improvements followed.

Nice To Haves

Experience developing platform services, infrastructure, or internal tooling in Go.
Experience building or owning a model registry, model hub, or similar system used by technical and nontechnical stakeholders.
Experience supporting ML systems where outcome labels are delayed or model quality cannot be measured immediately.
Track record of improving deployment speed, model quality, system reliability, infrastructure cost, or manual operational effort.
Experience designing configurable ML infrastructure that supports several modeling frameworks or approaches.
Evidence that reusable infrastructure, standards, or documentation were adopted across multiple teams.
Experience mentoring engineers or data scientists and influencing technical priorities across an ML portfolio.
Experience working with sensitive information and designing documentation or monitoring around appropriate access boundaries.

Responsibilities

Own the complete model lifecycle from experimentation through validation, deployment, versioning, retraining, rollback, and retirement.
Establish model promotion criteria, release processes, and rollback paths.
Design and build scalable ML data and feature engineering pipelines, including feature generation, historical aggregations, and training dataset construction, using Spark/PySpark and lakehouse technologies such as S3 and Apache Iceberg.
Develop incremental data processing pipelines with data quality validation, lineage, point-in-time correctness, and training-serving feature consistency.
Build monitoring for data quality, drift, prediction quality, service reliability, latency, and infrastructure cost.
Partner with data scientists to define offline and production evaluation signals, including for use cases with delayed outcome labels.
Use production evidence to improve model performance, platform reliability, deployment speed, and infrastructure efficiency.
Build reusable training, evaluation, and deployment pipelines for multiple models and modeling approaches.
Develop orchestration workflows for dependencies, scheduling, retries, backfills, and failure recovery.
Translate scientific requirements into maintainable production systems with clear interfaces, ownership, and operational expectations.
Build and maintain an accessible model hub connecting model documentation, evaluation results, release history, versioning, and monitoring.
Create tested CI/CD workflows, reproducible environments, actionable alerts, incident-response procedures, and operational runbooks.
Define MLOps priorities with Data Science and Engineering, identifying risks before they become production issues.
Establish reusable engineering standards, mentor teammates, and reduce the operational burden on data scientists.