Data Scientist

Big Shot Pictures•Los Angeles, CA

10d

About The Position

Data Scientist – Content Recommendations & Experimentation About Big Shot Pictures Big Shot Pictures is a next generation animation company founded by Brian Robbins that produces the franchises of tomorrow. With a distinctive digital-first approach, the company builds global animation franchises across all platforms – from digital to theatrical, gaming to streaming, and consumer products to experiences. Big Shot has a first-look theatrical distribution deal with Sony Pictures Entertainment and is backed by strategic investment from Greycroft, Sony Pictures, MarcyPen Capital Partners, ValueAct Capital, and CAA. The Role The Data Scientist designs and builds the recommendation and experimentation systems that power Big Shot's content strategy. Working closely with engineering, product, and creative teams, this role turns audience data into actionable programming decisions — what to make, how to package it, and how to validate what's working. The Data Scientist owns the full lifecycle from model development through testing, measurement, and iteration, ensuring our content decisions are driven by rigorous data science, not guesswork.

Requirements

4-7 years of experience in data science, machine learning, or applied analytics, with hands-on experience building recommendation or ranking systems end-to-end
Strong Python proficiency and comfort with ML frameworks (scikit-learn, PyTorch, TensorFlow, or similar)
Experience with vector embeddings and similarity search tools (FAISS or equivalent)
Solid foundation in experiment design - A/B testing, statistical significance, causal inference fundamentals
Experience building data pipelines and working with messy, real-world datasets
Actively uses AI tools (LLM APIs, AI coding assistants, etc.) as a regular part of your workflow - not just curious, but practiced
Strong communication skills; able to explain model behavior and recommendations to non-technical stakeholders in a fast-moving creative environment
Comfortable with ambiguity, ownership, and moving fast

Nice To Haves

Experience with unstructured and vector databases (e.g., Pinecone, Weaviate, Qdrant) and understanding of hybrid structured/unstructured database architectures is a plus
Familiarity with NLP techniques (topic modeling, text classification, semantic similarity) is a strong plus
Experience with YouTube, streaming, or media/entertainment data is a strong differentiator

Responsibilities

Design and build recommendation models that inform programming decisions across topics, formats, video length, scheduling, packaging, and more
Develop scoring and ranking systems that surface high-opportunity content strategies from audience and performance data
Define benchmarks with broader team and design systems to monitor and optimize based on insights
Build and maintain data pipelines that feed clean, reliable signals into recommendation models
Continuously improve model performance through monitoring, retraining, and incorporating new data sources
Design and own the studio's experimentation framework including A/B testing infrastructure, holdout groups, and measurement methodology
Define success metrics tailored to content performance: watch time, retention curves, audience growth, subscribe conversion
Build systems to capture content performance outcomes and feed them back into recommendation models, creating a continuous learning loop
Ensure experiments are statistically rigorous and results are trustworthy enough to drive real production decisions
Translate model outputs and experiment results into clear, actionable recommendations for creative and editorial teams
Build dashboards, reports, and lightweight tools that make data science outputs accessible to non-technical users
Partner with engineering to integrate recommendation outputs into studio workflows and internal tools
Act as a data science thought partner to leadership on content strategy and roadmap planning
Work within the studio's cloud infrastructure to deploy, monitor, and maintain ML models in production
Optimize for performance and cost efficiency across data pipelines and model serving
Document systems, methodologies, and runbooks so the team can operate and troubleshoot without bottlenecks