Machine Learning Engineer

HorizonAI Talent; San Francisco (CA); On-site; USD 125,000 - 250,000; Full time

Benefits offered by this job

Equity

Job summary

HorizonAI Talent is assisting a YC-backed startup building AI-powered CAD software for mechanical engineers. They are developing a foundation model for physical design.

This is a hands-on seat for an ML engineer who designs and trains their own models; it isn’t an agents, RAG or fine-tuning role. You will design and train custom models and embeddings for 3D and geometric data, work with transformer architectures (LLMs, diffusion, graph transformers) and graph neural networks, own the data

Qualifications

• 4+ years of industry ML experience or a PhD.
• A model you designed from scratch and trained yourself, including the training pipeline and the data work
• Model development in your current, full-time role (not only past research)
• Expert Python and PyTorch (JAX or TensorFlow also works)
• Hands-on experience with transformer architectures

Responsibilities

• Designing and training custom models and embeddings for 3D and geometric design data
• Transformer architectures (LLMs, diffusion, graph transformers) and graph neural networks
• Owning the data pipeline end to end: collecting, cleaning and preparing it yourself
• Building the training and evaluation infrastructure
• Taking models from prototype into production

Skills

Python

PyTorch

Transformer architectures

Graph neural networks

Tools

Python

PyTorch

JAX

TensorFlow

Job description

A client has asked us to find a Machine Learning Engineer: a YC-backed startup building AI-powered CAD software for mechanical engineers. They're building a foundation model for physical design. This is a hands-on seat for an ML engineer who designs and trains their own models. It isn't an agents, RAG or fine-tuning role.

What you'll work on:

• Designing and training custom models and embeddings for 3D and geometric design data
• Transformer architectures (LLMs, diffusion, graph transformers) and graph neural networks
• Owning the data pipeline end to end: collecting, cleaning and preparing it yourself
• Building the training and evaluation infrastructure
• Taking models from prototype into production

What you'll need:

• 4+ years of industry ML experience or a PhD
• A model you designed from scratch and trained yourself, including the training pipeline and the data work
• Model development in your current, full-time role (not only past research)
• Expert Python and PyTorch (JAX or TensorFlow also works)
• Hands-on experience with transformer architectures

Nice to have:

• Graph neural networks, world models, 3D/geometry, robotics or physical simulation
• MLOps and cloud experience

The details:

• San Francisco Bay Area (SM), 5 days a week onsite
• $125K to $250K base plus equity
• H-1B transfers and TN considered. The client can't sponsor a new H-1B