Warner Bros. Discovery

Welcome to Warner Bros.Discovery the stuff dreams are made of.Who We Are :When we say, the stuff dreams are made of, were not just referring to the world of wizards, dragons and superheroes, or even to the wonders of Planet Earth.Behind WBDs vast portfolio of iconic content and beloved brands, are the storytellers bringing our characters to life, the creators bringing them to your living rooms and the dreamers creating whats next.From brilliant creatives, to technology trailblazers, across the globe, WBD offers career defining opportunities, thoughtfully curated benefits, and the tools to explore and grow into your best selves.Here you are supported, here you are celebrated, here you can thrive.Senior Data Scientist (Video AI), Bangalore.About Warner Bros. Discovery :Warner Bros. Discovery, a premier global media and entertainment company, offers audiences the world's most differentiated and complete portfolio of content, brands and franchises across television, film, streaming and gaming.The new company combines Warner Medias premium entertainment, sports and news assets with Discovery's leading non-fiction and international entertainment and sports businesses.Meet Our Team :At Warner Bros.Discovery, we are reimagining how machine learning transforms storytelling.As part of the AI/ML organization, focusing on supporting applications of AI to video, the Machine Learning Engineer Services group powers infrastructure and backend services behind production workflows.We're looking for an experienced ML Engineer with strong fundamentals and infrastructure experience to help build reusable components and services for video understanding, video summary, and video classifications.You will be part of a team focused on re-training, model hosting, cost optimization, and managing production workflows at scale.Roles & Responsibilities :- Ship orchestrated pipelines for training, batch/stream inference, and evaluation with lineage, versioning, and observability.- Design, build, and operate reusable services for training, batch/online inference, and vector/lexical search that meet clear SLOs.- Build/own vector + lexical search stacks and indexes; manage ingestion, filters, caching, and re-ranking.- Build/own vector + lexical search stacks and indexes; manage ingestion, filters, caching, and re-ranking.- Create reliable data/feature and evaluation pipelines with strong lineage, versioning, and observability.- Implement model deployment/rollout patterns (canary/AB, shadow) with monitoring and fast rollback.- Drive cost, reliability, and performance improvements across GPU/CPU workloads.- Integrate and maintain vector search services and semantic similarity infrastructure.- Design scalable model serving solutions for open-source and foundation models.- Develop systems for experiment tracking, model versioning, and evaluation.- Monitor production models for drift and performance degradation.- Manage compute cost and resource optimization across distributed training jobs.- Integrate Human-in-the-Loop (HITL) workflows and offline labeling into training pipelines.- Support model deployment for varied model architectures, including Vision-Language Models, Convolutional Neural Nets, and Embedding Generation models.- Stand up and maintain Feature Store and data versioning infrastructure.- Architect and implement RAG pipelines for video metadata, summarization, and Q&A.- Build evaluation frameworks to assess LLM performance, hallucination frequency, and structured response accuracy.What to Bring :- 5+ years of experience in Applied science, with end-to-end ML workflow expertise.- Strong background in model retraining, fine-tuning, and evaluation techniques.- Research work and having publications/patents in some conferences.- Computer vision experience is a must for Applied Research role.- Suggest and develop models.- Experience owning production ML or data services end-to-end (design ? deploy ? operate) with strong debugging and on-call instincts.- Familiarity with modern orchestration and CI/CD for ML, containerized serving, and scalable storage/indexing.- Pragmatic approach to telemetry (traces, metrics, logs) and using it to improve quality, latency, and cost.- Strong collaboration with ML researchers/engineers and product teams.- Experience deploying and managing open-source model servers (e.g., Triton, TorchServe, Ray Serve).- Proficient in managing cost-effective distributed computing environments (e.g., Kubernetes, Ray, SageMaker).- Familiar with experiment tracking tools (e.g., MLflow, Weights & Biases) and model versioning strategies.- Experience with real-time inference systems and streaming data pipelines is a plus.- Familiarity with labeling tools, HITL workflows, and offline data curation strategies.- Comfort working in Agile development environments and collaborating across global teams.- Trains/finetunes video & image models, builds data pipelines, writes clean PyTorch/JAX, adds tests & ablations.- Practical experience with training curves, hyperparam search, model distillation/quantization, export to ONNX/TensorRT.- Familiar with modern detectors/segmenters (e.g., SAM/Mask2Former/Grounding-DINO, Retina/CenterNet/YOLOv8+), face detection and recognition models, embedding models.What We Offer :- A Great Place to work.- Equal opportunity employer.- Fast track growth opportunities. (ref:hirist.tech)