Principal Machine Learning Engineer

We are building the next generation of AI-driven game experiences by running generative AI models directly on players' devices—including phones, tablets, laptops, and desktops.

Our games run within a modern, browser-native runtime built on technologies such as WebGPU and WebNN. As a Principal Engineer – On-Device AI Inference & Systems, you will be the technical authority responsible for bringing state-of-the-art multimodal models—including transformers and diffusion networks—from research into production. Your work will focus on making these models fast, memory-efficient, and reliable while integrating them seamlessly into a production game engine.

This is a deeply hands-on, high-impact engineering leadership role. You will own the entire inference pipeline—from model export and optimization through low-level kernel tuning—to shipping AI-powered features that run at interactive frame rates within strict memory, power, and latency budgets.

You'll establish engineering standards, define runtime architecture, mentor senior engineers, and directly influence the AI experiences delivered to millions of players worldwide.

If you're passionate about bridging the gap between cutting-edge research and production software—and enjoy profiling, frame captures, operator fusion, and eliminating milliseconds and megabytes—this role is for you.

What You'll Be Doing

Inference & On-Device Optimization

• Own the complete optimization pipeline, including:

  • Model export

  • Graph transformations

  • Operator fusion

  • Memory layout optimization

  • Hardware-specific kernel tuning across NPUs, mobile GPUs, and desktop GPUs

• Make architectural decisions around:

  • INT4, INT8, and FP16 quantization

  • Weight sharing

  • Structured and unstructured pruning

  • Knowledge distillation

• Optimize models to meet aggressive latency, memory, and power targets while maintaining quality.

• Perform low-level GPU optimization by writing and tuning:

  • WebGPU compute shaders (WGSL)

  • Native compute kernels using Metal, Vulkan/SPIR-V, Direct3D 12, or CUDA

• Profile performance using tools such as:

  • Chrome/Dawn GPU traces

  • PIX

  • Apple Instruments / Metal System Trace

  • Snapdragon Profiler

  • NVIDIA Nsight

  • RenderDoc

• Apply advanced optimization strategies including:

  • Dynamic resolution

  • Token reduction

  • Cross-frame caching

  • Reduced-step diffusion samplers

Runtime & Systems Integration

• Evaluate and implement inference runtimes including:

  • ONNX Runtime Web

  • Transformers.js

  • WebLLM

  • TensorFlow.js

  • CoreML

  • TensorFlow Lite

  • ExecuTorch

• Extend or develop runtime infrastructure where existing solutions are insufficient.

• Design the integration between the ML runtime and the game engine, including:

  • Thread scheduling

  • Memory pooling

  • Zero-copy buffer sharing

  • Frame-budget management

• Architect real-time inference systems capable of processing:

  • Images

  • Text

  • Scene primitives

  • Metadata

• Build supporting infrastructure for:

  • Model packaging

  • Asset pipelines

  • Device capability tiers

  • Runtime fallbacks

  • Telemetry

  • Automated on-device benchmarking

Research Productionization

• Partner with research scientists to deploy new model architectures efficiently.

• Provide feedback to research teams regarding:

  • Hardware constraints

  • Operator support limitations

  • Performance cost models

• Evaluate advances in efficient inference, including:

  • Efficient attention

  • Distillation

  • Reduced-step diffusion

• Recommend engineering investments based on measurable improvements in latency, memory usage, and power consumption.

Engineering Leadership

• Lead and mentor a team of engineers.

• Establish engineering best practices, code review standards, benchmarking methodologies, and performance regression gates.

• Define and track KPIs for:

  • Latency

  • Quality

  • Memory usage

  • Power consumption

• Collaborate with platform engineering, runtime, and product teams to align AI capabilities with device constraints and product roadmaps.

What We're Looking For

• 8+ years of software or machine learning engineering experience.

• At least 4 years focused on:

  • On-device AI inference

  • Edge AI

  • Performance-critical real-time systems

• Proven experience shipping transformer or diffusion models on mobile, desktop, or embedded hardware.

• Hands-on experience deploying models with WebGPU using technologies such as:

  • ONNX Runtime Web

  • Transformers.js

  • WebLLM

  • TensorFlow.js

• Experience writing and optimizing WGSL compute shaders.

• Deep knowledge of at least one major inference runtime, including:

  • ONNX Runtime

  • ORT Web

  • CoreML

  • TensorFlow Lite

  • ExecuTorch

• Strong understanding of:

  • Operator fusion

  • Runtime scheduling

  • Memory layout optimization

• Low-level GPU programming experience using one or more of:

  • WebGPU / WGSL

  • Metal

  • Vulkan

  • Direct3D 12

  • CUDA

• Experience with model optimization techniques including:

  • INT4, INT8, and FP16 quantization

  • Weight sharing

  • Pruning

  • Knowledge distillation

• Familiarity with modern hardware platforms, including:

  • Apple Neural Engine

  • Qualcomm Hexagon

  • Adreno

  • ARM Mali

  • Apple Silicon

  • NVIDIA GPUs

  • AMD GPUs

  • Intel GPUs

• Strong proficiency in:

  • TypeScript

  • JavaScript

  • WGSL

  • Python

• Ability to understand, modify, and optimize deployed model architectures.

• Demonstrated technical leadership and mentoring experience.

Nice to Have

Experience with:

• On-device world models, neural rendering, or real-time generative AI, including:

  • NeRF

  • Gaussian Splatting (3DGS)

  • Real-time diffusion

• Game engine development using:

  • Unity

  • Unreal Engine

  • Custom rendering engines

• Graphics APIs including:

  • Metal

  • Vulkan

  • Direct3D

  • OpenGL ES

• Contributions to open-source inference runtimes or GPU libraries, particularly within the WebGPU ecosystem:

  • Dawn

  • wgpu

  • ONNX Runtime Web

  • Transformers.js

  • WebLLM

• Knowledge of emerging WebGPU features, including:

  • Subgroups

  • FP16 / shader-f16

  • Timestamp queries

• Compiler frameworks such as:

  • MLIR

  • TVM

  • IREE

  • XLA

• On-device benchmarking infrastructure, performance regression testing, and large device-farm automation.

Additional Information

• International relocation assistance is not available.

• Employment visa sponsorship is not available.

Compensation

Base Salary: $278,100 – $417,100 USD

Final compensation will depend on geographic location, experience, technical expertise, and overall qualifications.

In addition to base salary, this position may be eligible for:

• Equity awards

• Annual incentive compensation

• Other company bonus programs

Benefits

Unity offers a comprehensive benefits package designed to support employee well-being and work-life balance. Benefits vary by country and employment status but may include:

• Medical, life, and disability insurance

• Retirement and pension plans

• Employee stock ownership programs

• Commute subsidies

• Generous vacation and personal leave

• Parental leave and family care programs

• Mental health and wellness resources

• Employee Resource Groups (ERGs)

• Employee Assistance Program (EAP)

• Professional development and training

• Volunteer and donation matching programs

• Office snacks and workplace amenities

About Unity

Unity is the world's leading game engine, powering interactive experiences for more than 3 billion users every month. Beyond gaming, Unity enables organizations across automotive, manufacturing, healthcare, architecture, and other industries to create, simulate, and collaborate in immersive 3D environments.

Unity is committed to fostering an inclusive workplace and is proud to be an Equal Opportunity Employer.