About the job
As a Senior Software Engineer on the Foundation AI organization, you will sit at the epicenter of our foundation model efforts. While the research world is focused on architecture, you will be the architect of the data flywheel that makes VideoGen and 3DGen possible. You aren't just building pipelines; you are building the infrastructure that defines how our models perceive and generate virtual worlds in three dimensions and across time.
Responsibilities
High-Scale Data Orchestration: Architect and maintain automated pipelines for the ingestion, cleaning, and pre-processing of multi-modal datasets (video, 3D,) spanning petabytes of data
Synthetic Data Generation: Leverage image and video generation models to scale multi-modal synthetic datasets
Research-to-Production Bridge: Partner with research teams to create training data for research experiments – research and implement synthetic data creation pipelines
Scalable Evaluation Frameworks: Build and own evaluation—automating both heuristic-based metrics and human-in-the-loop interfaces to evaluate and benchmark training datasets and in-house foundation models
Model Deployment & API Architecture: Design and optimize high-throughput, low-latency Inference APIs for internal and external consumer access
Autonomous SOTA Tracking: Actively participate in literature reviews and paper reading groups to identify and implement the latest optimizations in generative modeling
Resource Efficiency & Observability: Implement monitoring pipeline health, optimizing data loading to ensure GPUs are used efficiently
Qualifications
Minimum
8+ years of experience as a research-focused data systems engineer (preferably working with 3D and video foundation models)
Expertise in building scalable ML data pipelines for both batch and real-time environments. Experience working with and processing very large datasets (Petabytes or more).
Python Proficiency: You can write high-quality Python code for automation, tooling, and infrastructure management
Experience with cloud data platforms and distributed processing technologies (e.g., Spark, Ray, Kubeflow, S3, etc.)
A Bachelor's degree or equivalent experience in Computer Science, Computer Engineering, or a similar technical field
Preferred
MLOps Experience: Knowledge of experiment tracking (Weights & Biases, MLflow) and versioning for massive datasets.
Custom Tooling Development: Experience building internal "human-in-the-loop" tools for data labeling specific to video or 3D.
C++ Knowledge: Optimize the performance of data loaders and being comfortable modifying engine code.
Game development and digital content creation tools: Experience with making Roblox games, using Blender, Unreal Engine, or Unity.