Sr. Engineering Manager, AI Runtime

Databricks · Mountain View, California; San Francisco, California · Executive Engineering - Pipeline · listed July 6, 2026

The shape of it

Seniority
Manager
Experience asked
3–8 years
Where
Not stated
Stated pay
$228,600 – $297,120 USD
Requirements listed
9
Length
781 words

In the posting’s own words

As a Senior Engineering Manager, you will lead the team owning both the product experience and the foundational infrastructure of AIR. You'll shape customer-facing capabilities while designing for scalability, extensibility, and performance of GPU training and adjacent areas, collaborating closely across the platform, product, infrastructure, and research organizations.

What it asks for · 9

  • 8+ years of software engineering experience, with 3+ years in engineering management.
  • Track record building and operating managed GPU training infrastructure at scale (100s/1000s GPUs).
  • Deep familiarity with distributed training frameworks (PyTorch, DeepSpeed, Composer, Megatron-LM) and parallelism strategies (FSDP, tensor/pipeline parallelism).
  • Experience with training resilience patterns: checkpointing, elastic training, and automated failure recovery for long-running jobs.
  • Understanding of GPU performance fundamentals including NCCL, interconnect topologies, and memory optimization.
  • Experience building platform products with clear SLAs where you've owned the customer experience, not just the backend.
  • Strong cross-functional leadership across platform, product, and research teams, with the ability to lead through ambiguity and deliver complex projects.
  • Excellent collaboration and communication skills across engineering, product, and research organizations.
  • BS/MS in Computer Science, Electrical Engineering, or related technical field.

What the job covers

  • Lead, mentor, and grow a high-performing engineering team responsible for the Custom Training product and its foundational infrastructure, including distributed training orchestration, cluster lifecycle, fault tolerance, and training efficiency.
  • Define and own the product and technical roadmap for AIR, balancing customer experience, functionality, and foundational investments.
  • Collaborate closely with product, research, platform, infrastructure teams, and customers to drive end-to-end delivery, from ideation and prioritization to launch and operation.
  • Drive architectural decisions and product design for managed GPU training at scale.
  • Advocate for customer needs through direct engagement, ensuring engineering decisions translate to clear product impact.
  • Build observability and reliability practices for long-running, multi-node training jobs, including checkpoint strategies, failure recovery, and operational runbooks.
  • Partner with recruiting to attract, hire, and develop top-tier engineering talent.

Degree language

  • BS/MS in Computer Science, Electrical Engineering, or related technical field.

Tools and skills named

Models & research
  • GPU6×
  • Deep learning
  • Fine-tuning
  • LLM
  • PyTorch
Data
  • Databricks2×
Cloud & infra
  • Observability
Frameworks
  • Node.js
Operations & finance
  • Recruiting
Product & design
  • Roadmap
Ways of working
  • Cross-functional

Words the posting leans on

  • training12×
  • product11×
  • engineering8×
  • infrastructure7×
  • customer6×
  • experience6×
  • platform5×
  • model4×
  • research4×
  • air3×
  • building3×
  • data3×
  • deep3×
  • gpu training3×
  • lead3×
  • build2×

Counted from the posting after the mission statement and the legal notices are set aside. The ones near the top are the ones a screener is looking for.

The posting, your resume, and the gaps between them. One click loads all three.

More open at Databricks

every open role at Databricks

How this page was made

An automated read of a public job posting, fetched August 25, 2026 and last changed by Databricks on August 18, 2026. Every list above is pulled from the posting’s own sentences — nothing rewritten, nothing added, no judgment about the role or the company. Counts and seniority are read off the text by rule, so they can be wrong where the posting is unusual. The original is the only thing that binds. Openings close without warning; check the source before spending an evening on it.