SumerSports

• Build and train CV models for sports video: player/ball detection, multi-object tracking, pose/keypoints, event/action recognition, identity association (re-ID). • Own the experimentation loop: hypotheses → ablations → error analysis → measurable improvements. • Design and maintain evaluation: task-appropriate metrics (e.g., MOT metrics, keypoint accuracy, event precision/recall), dataset slices, and failure taxonomy. • Improve data efficiency: augmentations, sampling strategies, handling label noise, weak/self-supervision where helpful. • Prototype and iterate on modern architectures (e.g., transformer-based detection/tracking, temporal models, multi-task setups). • Collaborate on dataset + labeling design: formats, schemas, tooling, versioning. • Help productionize models: packaging, batch/stream inference patterns, throughput/latency tradeoffs, robustness checks. • Add lightweight quality gates: reproducibility, automated eval, regression detection.

• Design and build AI-powered user features: Implement prompt-based and retrieval-augmented systems (RAG) that answer complex sports questions and generate insights. • Build agents and workflows that combine deterministic logic with LLM reasoning. • Prototype and iterate fast: Use prompting, tool orchestration, and retrieval design to rapidly build and refine AI behaviors. • Collaborate with Eval Engineers to build golden sets and automate quality checks before release. • Work closely with the LLMOps Platform team: Use shared eval frameworks, prompt registries, and model gateways. • Provide feedback loops to improve platform reliability, latency, and safety. • Integrate with deep learning and data systems: Combine structured stats, tracking data, and video-derived features into AI-powered applications. • Build APIs and UI layers that expose insights to coaches, teams, and partners. • Push the limits of applied AI: Explore how LLMs, retrieval, and deterministic logic can create novel sports analytics tools. • Use AI internally to accelerate development (code generation, testing, debugging).

• Build and operate robust data pipelines for ingestion, cleaning, and transformation using Databricks, Airflow, or Dagster • Develop efficient ETL/ELT workflows in Python and SQL to support both batch and streaming workloads • Collaborate with ML and AI teams to deliver high-quality datasets for training, evaluation, and production features • Model and maintain structured data assets (Delta, Parquet, Iceberg) for reliability, versioning, and lineage tracking • Implement orchestration and monitoring — schedule jobs, track dependencies, and automate recovery from failures • Ensure data quality and compliance through validation frameworks, schema enforcement, and audit logging • Contribute to data platform evolution — evaluate tools, standardize best practices, and improve developer experience • Support performance and cost optimization across compute, storage, and orchestration systems

• Build and operate the LLM Platform: Develop model routing, prompt registry, and orchestration services for multi-model workflows. • Integrate external LLM APIs (OpenAI, Anthropic, Mistral) and internal finetuned models. • Enable fast, safe experimentation: Implement automated evaluation pipelines (offline + online) with golden sets, rubrics, and regression detection. • Support CI/CD for prompt and model changes, with rollback and approval gates. • Collaborate cross-functionally: Partner with product pods to instrument RAG pipelines and prompt versioning. • Work with deep learning and data teams to integrate structured and unstructured retrieval into LLM workflows. • Optimize performance and cost: Profile latency, token usage, and caching strategies. • Build observability and monitoring for LLM calls, embeddings, and agent behaviors. • Ensure reliability and safety: Implement guardrails (toxicity, PII filters, jailbreak detection). Maintain policy enforcement and audit logging for AI usage.

• Design and operate ML infrastructure: Manage data, training, serving, and inference systems for high-throughput model workflows • Build scalable pipelines: Implement reproducible training and evaluation pipelines with versioning, scheduling, and artifact tracking • Optimize compute and cost: Tune GPU and CPU workloads, manage clusters, and drive efficiency via rightsizing, spot scheduling, and caching • Serve models in production: Operate APIs for low-latency inference with autoscaling, blue-green or canary rollouts, and rollback safety • Ensure reliability and observability: Define and own SLOs; instrument pipelines and services to track latency, cost, drift, and data quality • Secure and automate: Manage IAM, secrets, and container security; automate deployment pipelines via CI/CD and infrastructure as code • Collaborate cross-functionally: Partner with research scientists and AI engineers to deliver models from experiment to production with minimal friction • Document and enable: Build templates, runbooks, and internal tooling that make ML workflows repeatable, safe, and fast

• Build and implement analytical football models, providing valuable insights to enhance our football analytics offerings • Conduct comprehensive data analysis to extract actionable insights • Ensure seamless deployment of machine learning and statistical models using Python, Databricks, and Spark • Handle large datasets efficiently using Spark and Databricks, ensuring data integrity and performance optimization • Work closely with other data scientists, analysts, and stakeholders to understand their needs and deliver high-quality analytical solutions • Identify and implement new tools, techniques, and best practices to enhance our data analytics capabilities