Senior Machine Learning Engineer, Discovery Recommendations

About the role

You will architect and own the end-to-end discovery recommendation systems that define how players find content inside Fortnite's main hub. You will lead the design of candidate generation and ranking models that must perform at scale for one of the largest live player bases in interactive entertainment. You will collaborate with data scientists, analytics engineers, and content quality teams to define and measure the success of recommendation quality and business impact. You will implement deep learning approaches such as two-tower architectures, transformer-based sequence models, and embedding-based retrieval to model user intent and content semantics. You will balance relevance, diversity, and fair exposure while navigating a rapidly changing catalog where new experiences launch daily and feedback loops are strong. You will drive experimentation culture by designing, running, and analyzing A/B tests that move from insight to production deployment. You will make pragmatic infrastructure tradeoffs between batch, near-real-time, and streaming serving to meet latency and accuracy targets. You will bring a passion for gaming contexts and an understanding of how recommendation design influences player experience and creator outcomes.

Key facts

What you'll do

Design and implement retrieval, ranking, and reranking models for creator content using deep learning approaches such as two-tower architectures, transformer-based sequence models, and embedding-based retrieval, while building user representation systems that power personalized discovery across the Epic catalog.

Build and optimize multi-stage candidate generation and impression allocation pipelines that balance relevance, diversity, and fair content exposure across a large and rapidly evolving catalog of creator-built experiences.

Design and run A/B experiments to validate model improvements, own evaluation frameworks that capture recommendation quality holistically, and drive the path from experiment to production deployment with strong observability.

Collaborate with analytics and content quality teams on ranking signals including genre classification, creator credibility, and content quality metrics to improve the fidelity of recommendations.

Own ML infrastructure decisions for the discovery stack, choosing the right tradeoffs between batch, near-real-time, and streaming serving architectures to meet accuracy and latency goals.

Explore and implement strategies for content cold-start and counterfactual evaluation methods, including explore/exploit approaches, to ensure new experiences can surface and learn quickly with sparse initial signals.

Apply content understanding models using NLP, computer vision, or generative AI features as inputs to ranking and retrieval systems, improving matching between user intent and catalog items.

Work in a cloud-based ML environment leveraging PyTorch, TorchRec, Transformers, Ray, Databricks, and AWS to deliver scalable, robust, and maintainable recommendation solutions.

Partner with cross-functional stakeholders to align recommendation strategy with product goals, creator incentives, and quality standards in the dynamic gaming and creator economy.

Demonstrate strong Python engineering skills to prototype, iterate, and productionize models and pipelines that handle massive scale and real-world data dynamics.

Requirements

5+ years of experience building production recommendation or ranking systems, ideally in a UGC, marketplace, or content discovery context, with a track record of deploying models that influence user behavior at scale.

Experience with deep learning for information retrieval and multi-stage recommendation pipelines, including candidate generation, scoring, and reranking, using architectures such as two-tower models and transformer-based sequence models.

Demonstrated ability to design and analyze A/B experiments, with awareness of biases inherent to recommendation systems, selection bias, and evaluation metrics that reflect long-term user and creator outcomes.

Strong Python engineering skills with experience in PyTorch and large-scale data processing frameworks such as Spark, including writing maintainable, tested, and performant code.

Comfort working in a cloud-based ML environment on platforms such as AWS, with experience managing data pipelines, model training workflows, and serving infrastructure.

Experience with explore/exploit strategies, content cold-start challenges, and counterfactual evaluation methods applied to recommendation systems to handle sparse initial data and evolving catalogs.

Experience with content understanding models including NLP, vision, or generative AI techniques used as features in ranking and retrieval systems to capture semantic similarity and intent.

Familiarity with creator economy dynamics and how recommendation design affects content quality, creator incentives, diversity of experiences, and ecosystem health.

Experience with the machine learning stack including PyTorch, TorchRec, Hugging Face Transformers, Ray, Databricks, and AWS services used for data, training, and serving.

Passion for video games and/or experience with gaming analytics, including understanding of player engagement, retention, and monetization dynamics in live service environments.

Nice to have

Experience building and operating real-time serving systems for recommendations in high-traffic, low-latency environments.

Background in recommender systems research, open source contributions, or publications in relevant ML venues.

Experience with fairness, explainability, and bias mitigation techniques specific to recommendation systems.

Knowledge of gaming platforms, live operations, and how recommendation design intersects with player experience and monetization.

Experience with MLOps practices, model monitoring, and experimentation infrastructure to ensure reliable and safe deployment of recommendation models at scale.

Practical notes

This role is open to multiple locations across the US (including CA, NYC, & WA).

Employment terms and specific location details are defined in the source engagement information.

No additional benefits or compensation details are specified in the source material beyond the high-level benefits overview.