5+ years of experience building or optimizing production AI systems.
Strong understanding of inference performance across compute, memory, and storage architecture.
Hands-on experience at the systems layer, specifically with GPU and CPU resource management.
Demonstrated ownership in model serving, retrieval, caching, storage, or distributed performance.
Ability to bridge architectural decisions and hands-on implementation in efficiency-driven environments.
Background in technically demanding fields like AI infrastructure or high-performance systems is essential.
PhD preferred but real-world experience is more valuable.
Responsibilities
Build and optimize LLM serving and inference systems for production.
Enhance performance on GPU and CPU pathways.
Address KV cache, memory, storage, and throughput bottlenecks.
Design and scale systems for retrieval-heavy AI workloads.
Contribute to infrastructure that improves AI performance through storage architecture efficiencies.
Resolve engineering challenges at the intersection of AI, high-performance systems, and distributed infrastructure.
Benefits
Work on the infrastructure that ensures AI systems are fast and scalable.
Engage with deep systems problems at the mechanical level of AI.
Opportunity to impact how AI performance translates commercially.
Collaborate with a team prioritizing performance, scale, and architecture in AI.
Be part of cutting-edge advancements in AI infrastructure.
Full Job Description
What you'll do
Build and optimize LLM serving and inference systems for production environments
Improve performance across GPU and CPU pathways
Work on KV cache, memory, storage, and throughput bottlenecks
Design and scale systems that support RAG and retrieval-heavy AI workloads
Contribute to infrastructure where storage architecture and systems efficiency materially affect AI performance
Solve engineering problems at the intersection of AI, high-performance systems, and distributed infrastructure
What we're looking for
An engineer who has spent meaningful time building or optimizing production AI systems, not just experimenting with models
Someone who understands how inference performance is shaped by the interaction between compute, memory, storage, and serving architecture
Deep hands-on experience working close to the systems layer - for example, improving how workloads run across GPU and CPU resources, reducing bottlenecks, or tuning infrastructure for better throughput and latency
Evidence of real ownership in areas like model serving, retrieval, caching, storage, or distributed performance, rather than purely application-layer AI work
The ability to move comfortably between architecture decisions and hands-on implementation, especially in environments where efficiency and scale matter
A background that suggests you can operate in technically demanding environments, whether that comes from AI infrastructure, high-performance systems, storage platforms, or adjacent distributed systems work
PhD preferred, but far less important than having built serious systems in the real world
Why this role is compelling
This is not a "prompt engineering" job.
This is not an "AI wrapper" job.
This is not a generic backend role with AI sprinkled on top.
This is a chance to work on the infrastructure that determines whether modern AI systems are fast, scalable, efficient, and commercially viable.
If you want to work on the real mechanics of AI performance - serving, retrieval, compute efficiency, memory behavior, storage architecture, and inference at scale - this is where that work happens.
Who will love this role
Engineers who enjoy deep systems problems
Builders who care about performance, scale, and architecture
People who want to work where AI meets infrastructure
Candidates who would rather solve hard technical bottlenecks than ship surface-level AI features
Who should not apply
This role is not for:
Purely academic researchers without meaningful production ownership
Generic software engineers without clear AI systems or inference depth
Candidates focused mainly on prompt engineering or lightweight application integrations
MLOps generalists who have not worked deeply on serving, storage, or performance-critical AI systems
