Apply on Employer Site

Mirage · 1 month ago

Member of Technical Staff, Training Data Infrastructure

New York, NY

Full-time

Onsite

Mid, Senior Level

$215K/yr - $300K/yr

3+ years exp

Mirage is the leading AI short-form video company, building full-stack foundation models and products for video creation. The role involves developing training data infrastructure and optimizing data processing systems for multimodal datasets, directly impacting the company's ability to train models for millions of users.

Artificial Intelligence (AI)Generative AISoftware

Responsibilities

Build performant pipelines for processing video and multimodal training data at scale

Design distributed systems that scale seamlessly with our rapidly growing video and multimodal datasets

Create efficient data loading systems optimized for GPU training throughput

Implement comprehensive telemetry for video processing and training pipelines

Create foundation data processing systems that intelligently cache and reuse expensive computations across the training pipeline

Build robust data validation and quality measurement systems for video and multimodal content

Design systems for data versioning and reproducing complex multimodal training runs

Develop efficient storage and compute patterns for high-dimensional data and learned representations

Own and improve end-to-end training pipeline performance

Build systems for efficient storage and retrieval of video training data

Build frameworks for systematic data and model quality improvement

Develop infrastructure supporting fast research iteration cycles

Build tools and systems for deep understanding of our training data characteristics

Build infrastructure enabling rapid testing of research hypotheses

Create systems for incorporating user feedback into training workflows

Design measurement frameworks that connect model improvements to user outcomes

Enable systematic experimentation with direct user feedback loops

Qualification

ML infrastructure developmentDistributed computing frameworksHigh-throughput data pipelinesVideo/image data processingMachine learning workflowsPerformance optimizationCluster managementMLOpsTeam collaborationProblem-solving