AI Engineer - Genius Sports - New York, United States

About the Role - AI Engineer, Sports AI We're looking for an AI Engineer on our Sports AI team to help build the next generation of applied AI systems powering sports analysis, automation, and insights. These systems use live and historical sports data, including tracking data, structured feeds, broadcast video, commentary, and text, to understand game context, detect & enrich key events, estimate the probability of future events, and generate insights. The outputs from these systems power a range of products and workflows, such as projecting which games or moments will be most exciting to fans and automating parts of manual play-by-play collection using CV/AI. The role spans a broad set of sports modeling and automation problems across multiple sports, including soccer, American football, and basketball. This role sits at the intersection of machine learning, AI system design, and production engineering. You'll own scoped AI systems end-to-end: framing the relevant modeling problems, constructing the datasets needed to solve them, training and composing models and algorithms, building the inference pipelines that orchestrate them, and rigorously evaluating output quality against messy, real-world data. You'll work on challenges like aligning signals across multiple sources, handling uncertainty and inconsistency in system outputs, and improving accuracy, latency, and reliability in real-time production workflows. In this role, hands-on ML/AI work will be central: understanding data, developing models and algorithms, evaluating outputs empirically, and iterating in production. You'll also apply LLMs and agentic workflows as part of your broader AI engineering toolkit. Key Responsibilities - Own applied AI work end-to-end, from data exploration and early prototypes through evaluation, production integration, and iteration - Develop and compose models, algorithms, and inference pipelines that convert sports data into structured events, predictions, insights, and confidence-aware outputs - Build models for problems such as event detection, event likelihood estimation, fan interest & excitement projection, and automation of manual play-by-play collection - Work with messy, multimodal sports data from tracking systems, video and computer vision outputs, audio, commentary, text, and structured feeds, including imperfect labels and ambiguous real-world examples - Define and use metrics, evaluation datasets, and benchmarks to measure AI system quality and guide model, algorithm, and product decisions - Train, adapt, evaluate, and integrate ML models and AI components, including multi-step systems where model, algorithmic, and LLM/agent outputs are composed, validated, and refined - Design workflows that use human review or correction data to improve evaluation, model iteration, and production output quality where appropriate - Work closely with CV engineers on training pipelines, labeling workflows, and model deployment patterns - Partner with product, data platform, infrastructure, and systems engineers to integrate evaluated AI outputs into real-time sports products and automation workflows - Mentor junior teammates and contribute to team knowledge-sharing, reviews, and experiment design Qualifications - 3+ years of experience building production ML, CV, or AI systems - Ability to translate ambiguous sports product goals into concrete ML tasks, including defining the prediction target, identifying the right data, measuring output quality, and shipping production-ready solutions - Hands-on production ML/AI experience, including constructing datasets, defining features and labels, training and deploying models, evaluating outputs empirically, and shipping AI system capabilities into production - Strong modeling judgment across deep learning and classical ML, with experience choosing approaches based on data inputs and problem structure - Experience with predictive modeling, event detection, data labeling, data quality improvement, and communicating experiment results to technical and non-technical stakeholders - Ability to evaluate AI system quality beyond anecdotal inspection, including reasoning about ambiguous outputs, imperfect labels, uncertainty, and real-world product tradeoffs - Strong production engineering fundamentals, including testing, observability, performance, and reliability - Demonstrated interest in the fast-moving landscape of LLMs, latest models, agentic AI systems, and development frameworks - Comfortable working in fast-moving, iterative environments with evolving requirements Preferred Qualifications - Hands-on experience with LLM-integrated workflows, LLM APIs or cloud AI platforms such as AWS Bedrock, agentic AI systems, multi-agent systems, or evaluation of LLM/agent outputs in production workflows - Experience with ML/CV domains relevant to sports understanding, such as action recognition, sequence modeling, multimodal modeling, object detection, tracking, or player identification - Experience working with player tracking data, sports analytics, play-by-play data, labeling platforms, and/or ML training platforms such as Union - Experience collaborating with CV engineers or integrating CV model outputs into downstream ML workflows - Experience using human review or correction workflows to evaluate and improve AI system quality - Experience building production systems in Rust - Familiarity with streaming, event-driven, audio/video, or real-time data workflows is a plus - Background or strong interest in sports, especially soccer, American football, and basketball