Generative AI - ML System Engineering

Department: Research

Location: Sunnyvale

Employment Type: FullTime

Who You Are

We are looking for Machine Learning Systems Engineers who can help us build the world's largest end-to-end 3D native machine learning systems. You will help us build our end to end ML framework dedicated for 3D, from pretraining, to finetuning, inferencing, etc. We expect a combination of strong hands on engineering skills, eagerness to learn new things, and thrives in a fast-paced, high-ownership environment.

Who We Are

At Meshy, we believe 3D creation should be boundless and accessible. Our mission statement is simple: unleash creativity. We built a full pipeline for 3D content ranging from text / image to 3D, texturing, texture editing, animation rigging, etc. We also built a vibrant community for our creators, where people can share their work, take inspiration from others, and even use it as an asset marketplace for their games and prototypes. We are the market leader in 3D generative AI, recognized as the No.1 in popularity among 3D AI tools (according to 2024 A16Z Games survey), and we generate real value and is used by enterprises (including Meta, Square Enix, Deepmind, etc.) and millions of end users. Meshy is used in game and film production, in 3D printing, in industrial product design, in enablement of novel product features such as user-generated content, and even in training and simulation for robotics and physical AI.

Your next challenge

3D is the brave new frontier of Gen AI. Our work here involves a lot of unique new challenges in both training and inference. Your next challenge at Meshy would involve the full stack of AI, from debugging and monitoring the hardware platform, building training framework, scaling high-throughput 3D data pipelines for our foundational training, co-designing novel model architectures with researchers, to the novel challenge of efficient inference engines for diffusion models and more. Here are some examples for each side of the challenge:

On the training side

• Work closely with researchers to co-design the next frontier of 3D & Spatial AI.

• Build and debug on top of modern PyTorch, for maximum parallelism and efficiency, and build clean and intuitive training infrastructure for our in-house foundational models.

• Identifying bottlenecks and optimizing for high throughput & efficient distributed model training across hundreds to thousands of GPUs.

• Implementing and maintaining 3D specific custom operators in Triton or CUDA.

• Implementing and maintaining novel data-loading framework and libraries.

On the inference side

• Building efficient inference endpoints with complex multi-stage model pipelines.

• Optimizing models through compilation, fusion, quantization, etc.

What We're Looking For

• Experience in machine learning or high performance graphics.

• Solid practical understanding of at least one machine learning framework (e.g. PyTorch, JAX).

• Strong ability to write beautiful and maintainable code in Python and/or C++.

• Ability to learn fast and dive into new concepts or complex codebases.

• Performance and efficiency oriented mindset, with a strong interest in the tiniest detail.

• Strong communication skills for working in a globally distributed team.

Nice to have

• A strong passion to navigate through the PyTorch internals, with hands-on experience in areas like torch.compile , fully_shard (FSDP2) APIs.

• Experience with building Triton kernels.

• Experiences with large-scale distributed training, familiarity with modern parallelization techniques: DP, TP, CP, PP, zero redundancy optimizers, etc.

• Experience with diffusion models in 3D or video.

• Experience with low precision bf16 or fp8 training.

A Little More about meshy.ai

Trusted by Meta, Square Enix, Deepmind and more, Meshy is redefining 3D creation with generative AI. We empower artists, designers, engineers, hobbyists, and makers to bring immersive worlds, characters, and experiences to reality in minutes instead of months.

In addition to our core mission of unleashing creativity, we build a culture that we enjoy and are proud of. Here are some highlights:

• We value intelligence and the pursuit of knowledge. We are a global team of generative-AI pioneers, computer-graphics veterans, and product builders who believe human expression and enjoyment is the ultimate frontier of computing.

• We care deeply about our work, our users, and each other. Empathy and passion drive us forward. We have a culture of directness and truthfulness, therefore we value constructive criticism. Being direct and truthful is the most sincere form of trust and care.

• We trust our instincts and are not afraid to take bold risks. Meshy was born from a few-hour prototype, a bold pivot for a team that had very little experience in AI. Innovation requires courage.

• We have a keen eye for quality and aesthetics. Our products are not just functional but also beautiful. The same aesthetics permeate through our culture, our code and are the same: functional and beautiful.

Locations and work environment

Persons in these roles preferably should be able to spend part of their time on-site in our Sunnyvale HQ. On-site requirements vary based on position and team, and we support fully remote options for exceptional candidates. We also have employees in Seattle, Boston, NY, Toronto, etc. If you have questions about hybrid or remote work arrangements for this role, please ask your recruiter.

We are a fast paced startup and we work hard. You need to be able to:

• communicate information and ideas so others will understand.

• observe details at close range.

• work under deadlines.

Interview process

1. You will first be contacted by our recruiter.

2. Soon afterwards, you will receive an online assessment of your knowledge about various engineering topics around training, inference, transformer architecture, and simple numpy coding exercises.

3. We will then schedule a 45 minutes - 2 hr interview slot for a technical coding round. The questions will revolve around performant C++ programming, tensor / array programming in PyTorch, and some practical hands-on open-book / open-internet training exercises in our GPU-enabled jupyter notebook.

4. Finally, you will be invited for a 3 hr onsite interview where we'd like you to present a previous work that you are proud of, then you will demonstrate your debugging skills and performance sense in a session with one of our engineers. Finally, you will talk to one of our leaders and our CEO about our culture, your background, and whether we have matching vibes.