Principal Machine Learning Engineer, Presentation & Visual Optimization

Paramount•New York, NY

1d•$139,000 - $180,000

About The Position

We are seeking a Principal Machine Learning Engineer to lead our Presentation pod. While other teams within AMLG focus on the "recommender" (the selection of content), your pod owns the "visual gateway." Your mission is to optimize how content is displayed to capture user attention and communicate value instantly. You will own the machine learning strategy for artwork selection, marquee personalization, carousel design, and title optimization across our global streaming platforms. This role is at the intersection of Machine Learning, Product Design, and Visual Psychology. You will lead a high-velocity pod that specializes in Multi-Armed Bandits (MAB)and computer vision to identify which image, headline, or layout style resonates most with a specific user in a specific context. This is not about recommending a movie; it is about ensuring that once a movie is recommended, it is presented in the most compelling way possible to drive a click. The Presentation pod has one of the highest "surface-to-impact" ratios in the company. In this role, you will directly shape: The First Impression: Owning the "Artwork Personalization" engine that determines the visual identity of our catalog for every user. Attention Optimization: Transitioning beyond standard recommendation to solve for "visual fatigue" and "discovery friction" through dynamic layout and title changes. Experimentation at Scale: Leading a pod built for swiftness, running hundreds of concurrent bandit-based experiments to find the winning visual combination in real-time.

Requirements

6-8+ years of experience in machine learning engineering or applied science.
Bandit Expertise: Deep hands-on experience with Multi-Armed Bandits, Contextual Bandits, Thompson Sampling, or Upper Confidence Bound (UCB) algorithms.
Rapid Experimentation: Proven track record of designing high-velocity A/B testing or online learning systems.
Technical Stack: Proficiency in Python, PyTorch/TensorFlow, and big-data processing (Spark/Databricks).
Leadership: Experience leading a technical pod or team, with a focus on translating product/design needs into engineering requirements.

Nice To Haves

Experience in Visual Personalization (artwork, thumbnails, or creative optimization).
Background in Computer Vision (OCR, aesthetic scoring, or image feature extraction).
Knowledge of Reinforcement Learning (RL) for sequential layout optimization.
Familiarity with LLMs/Generative AI for automated copy generation and title testing.

Responsibilities

Lead the Presentation Pod: Define the technical roadmap for visual personalization, bridging the gap between ML science and UI/UX design.
Artwork & Marquee Personalization: Architect and deploy multi-armed bandit (MAB) systems to dynamically select the best creative assets for show tiles and hero marquees.
Layout & Style Optimization: Develop models to personalize carousel design types (e.g., square vs. poster), presentation styles, and even the ordering of attributes shown to the user.
Title & Copy Optimization: Utilize NLP and LLMs to experiment with and optimize carousel titles and content descriptions based on user preferences and trending topics.
Fast Experimentation Frameworks: Design "always-on" experimentation pipelines that can shift traffic based on model confidence and reward signals (CTR, Playback Start).
Visual Understanding: Partner with Content Engineering to leverage visual embeddings and computer vision signals to understand why certain artwork performs better than others.