Machine Learning Engineer

About Owl AI

Owl AI is the world's first end-to-end AI platform built to bring fairness, clarity, and immersion to judged and referee-influenced sports. Founded by Olympic athlete and X Games CEO Jeremy Bloom alongside former Google Cloud AI Chief Josh Gwyther, we made history at the 2025 X Games Aspen becoming the first artificial intelligence to judge a professional sport on live broadcast television.

Backed by S32, Menlo Ventures, and Susa Ventures, and trusted by leagues such as Major League Pickleball, Owl AI is rapidly expanding into new sports and applications—from real-time officiating and line-calling to deep scouting analytics and multilingual live commentary.

We are a small team of high-conviction technologists on a mission to modernize how sports are experienced, judged, and understood.

We are looking for an experienced applied ML engineer to join our small but mighty engineering team. This is a high-ownership role where you'll work directly with our applied research group and engineering leadership to build the machine learning systems that power our live sports officiating, advanced analytics, and emerging media applications.

You thrive in ambiguity and bring a researcher’s rigor with a builder’s bias for shipping. You're as comfortable fine-tuning a foundational model or improving a detection pipeline as you are building the infrastructure that trains, deploys, versions, and monitors them in production. From experimentation through live inference, you will own ML problems end-to-end and build the operational backbone that makes them dependable.

Applied ML & Modeling

• Design, train, fine-tune, and evaluate machine learning models across various domains (spanning foundational models, generative AI, computer vision, and predictive analytics) for real-time officiating and sports applications.

• Build evaluation frameworks that hold models to broadcast and officiating standards, ensuring outputs are accurate, defensible, and robust.

• Adapt state-of-the-art research into practical, production-ready solutions tailored to our unique latency and accuracy constraints.

ML Ops & Production Systems

• Own the deployment path for models: establishing CI/CD for training and inference, as well as robust model versioning and registries.

• Build and maintain data, labeling, and training pipelines with experiment tracking and reproducibility to ensure results are traceable.

• Stand up monitoring for models in production—tracking performance, drift, and data-quality signals to catch regressions before they impact live broadcasts.

• Optimize models to run efficiently under real-time latency budgets, partnering with platform engineers on scalable serving and integration.

Operational Excellence

• Design for reliability, observability, and reproducibility from day one.

• Champion sound experimental and engineering practices within a collaborative, fast-moving environment.

• Continuously explore and adopt new models, techniques, and AI-assisted workflows to raise both model quality and team velocity.

Required

• Five to ten years of engineering experience, spanning applied machine learning and the infrastructure that supports it.

• Deep proficiency in Python for model development, data processing, and production ML code.

• Hands-on experience training and deploying models in PyTorch, TensorFLow, JAX, or similar frameworks.

• Demonstrated experience taking ML models from experimentation to production, including training, evaluation, deployment, and ongoing operation.

• Experience building and maintaining production systems and data pipelines.

• Comfortable in a small team environment: collaborative, self-directed, and highly accountable.

• Strong and proactive communication; you surface technical tradeoffs, risks, and opportunities clearly.

Nice to Have

• Experience fine-tuning open-weight models for specific downstream tasks.

• Familiarity with Google Cloud Platform (GCP) and its ecosystem of data and ML tools.

• Experience building hardware-accelerated video processing and streaming pipelines using GStreamer, NVIDIA Deep Stream SDK, and CUDA.

• Familiarity with model optimization and high-performance inference engines,…