Latest AI JobsBrowse Jobs
World Labs logo

Performance Engineer (Inference, Training & GPU)

World LabsSan FranciscoEngineering
Apply on company site
Salary
$200K - $300K USD
Job type
Full-time
Posted
1 May 2026
Status
Open · checked 7d ago

About this role

World Labs is hiring a full-time Performance Engineer in San Francisco to accelerate spatial intelligence models for maximum hardware efficiency. This hands-on, individual-contributor role suits a technical specialist ready to tackle complex systems challenges at a pioneering artificial intelligence startup.

Role Overview

As a Performance Engineer, the primary mission is making large generative world models train and serve as quickly as hardware allows. This involves finding and eliminating bottlenecks across kernels, serving paths, training loops, and hardware utilization. The position requires working across the entire stack, spanning low-level tensor optimization and fleet-wide serving efficiency, all while maintaining strict numerical correctness.

Key Responsibilities

Optimize inference and serving pipelines end to end, focusing on latency, throughput, batching, caching, and scheduling for production scale.
Write and tune custom GPU kernels using CUDA and Triton for hot paths, driving kernel fusion, bandwidth-bound optimization, and low-precision execution.
Enhance training throughput and GPU utilization through advanced parallelism strategies, communication and compute overlap, mixed precision, and eliminating pipeline stalls.
Build robust performance models, profiling workflows, and observability tools to track throughput, latency, cost, and utilization tradeoffs.
Maintain numerical correctness across precision, kernel, and hardware modifications.
Collaborate closely with researchers to productionize models for serving and accelerate experimental workflows.
Work on distributed systems supporting training and inference workloads when necessary.

Requirements

Strong performance-engineering fundamentals including profiling, roofline analysis, and root-cause investigation.
Deep GPU programming and optimization experience with CUDA or Triton.
Hands-on background optimizing inference and serving for large models, such as batching, KV or prompt caching, quantization, and low-latency sampling.
Practical experience optimizing training performance through parallelism, distributed communication, and mixed precision.
Working knowledge of machine learning framework internals like PyTorch, JAX, torch.compile, or XLA.
Strong proficiency in Python with the ability to write C++, CUDA, Rust, or Go code as needed.

Nice to Have

Background at an AI lab or machine learning native company optimizing systems for researchers.
Expertise in low-precision numerics including FP8 and INT8 quantization.
Experience with distributed systems for large-scale training and inference, such as NCCL and NVLink.
History of serving generative, diffusion, video, or 3D and spatial models beyond text language models.
Multi-accelerator experience involving GPUs, TPUs, or Trainium.
Experience building performance-modeling and observability frameworks for GPU utilization.

Skills and Tools

Python
PyTorch
Rust
C++
CUDA
Triton
JAX
XLA

Salary and Benefits

Annual salary range between $200,000 and $300,000 USD.
Generous health, dental, and vision insurance.
401K retirement benefits.
Flexible paid time off and paid parental leave.
Relocation support provided as needed.
Visa sponsorship available for qualified candidates.

About the Company

World Labs is a frontier AI research and product company advancing spatial intelligence, co-founded by industry leaders including Dr. Fei-Fei Li, Justin Johnson, and Ben Mildenhall. The organization pioneers world models that perceive, generate, reason, and interact with virtual and physical spaces. Its flagship product, Marble, transforms text, images, and video into navigable 3D worlds for gaming, film, architecture, and robotics. Backed by prominent investors with over $1 billion raised, the company is building an in-person team in San Francisco.

How to Apply

Interested candidates should visit the official World Labs careers page to submit an application for this position.

Skills and tech stack

PythonPyTorchRustC++

Questions candidates ask

Where is the Performance Engineer position located?

The role is based in San Francisco, California, working in the office 5 days a week.

What is the salary range for this job?

The annual salary is between $200,000 and $300,000 USD.

Does World Labs offer visa sponsorship?

Yes, World Labs sponsors visas for qualified candidates.

What programming languages are required?

Strong proficiency in Python is required, along with the ability to write C++, CUDA, Rust, or Go as needed.

What benefits are included with employment?

Benefits include health, dental, vision, 401K, flexible PTO, paid parental leave, and relocation support.

Ready to apply?

The application is on the official World Labs careers page.

Apply on company site