Senior ML Infra Architect: Scalable Data Pipelines

Adobe Inc.; San Jose (CA); On-site; USD 268,000 - 388,000; Full time

Job summary

Adobe Inc. in California seeks a Principal ML Engineer to architect and scale multimodal data pipelines for Firefly’s foundation models. You will build distributed, GPU-accelerated systems turning billions of raw assets into training-ready data at scale.

As a senior IC, you’ll influence data, infrastructure, and modeling teams, drive throughput and data curation, and ensure reliability while enabling rapid model learning progress.

Qualifications

• 10+ years in data engineering, ML infra, or distributed systems at large scale.
• Strong software engineering background with distributed systems and Ray/Spark experience.
• Proficiency in Python and systems languages (C++, Rust, Go, or Java).
• Deep knowledge of scalable databases and data storage.
• Experience optimizing GPU inference for VLMs/LLMs and data curation for training.

Responsibilities

• Architect and optimize large-scale distributed pipelines processing billions of assets into training-ready data.
• Scale inference throughput across pipelines and optimize hardware utilization.
• Design and own architecture for databases, storage, and GPU/CPU workloads.
• Collaborate with modeling teams to translate data requirements into pipelines.
• Bridge data engineering and applied ML across multiple teams.

Skills

Distributed systems

Ray/Spark

Python

Systems programming

Databases and data lakes

GPU inference optimization

Data curation for model training

Cross-team collaboration

Debugging distributed systems

Education

Bachelor's/Master's/PhD in CS/Engineering/ML

Tools

Ray

Spark

Job description

Adobe Inc. in California seeks a Principal ML Engineer to architect and scale multimodal data pipelines for Firefly’s foundation models. You will build distributed, GPU-accelerated systems turning billions of raw assets into training-ready data at scale.

As a senior IC, you’ll influence data, infrastructure, and modeling teams, drive throughput and data curation, and ensure reliability while enabling rapid model learning progress.