ML Engineer, Research

Quiet Capital; San Francisco (CA); On-site; USD 150,000 - 190,000; Full time

Job summary

Mach9 seeks an ML Engineer to build perception models for their AI-enabled CAD system. You will develop 3D and 2D detection from LiDAR and imagery, moving from research prototypes to production features with cloud integration in Digital Surveyor.

The role blends research with product delivery, requiring ownership from architecture papers to shipped capabilities and collaboration with infrastructure and product teams.

Qualifications

• Strong foundation in computer vision and deep learning.
• Experience taking an ML model from research to production.
• Proficient with Python and ML libraries (PyTorch/JAX/TF).
• Strong communication and collaboration skills.

Responsibilities

• Design, train, and evaluate CV and 3D ML models for CAD-grade geometry and features from LiDAR and imagery.
• Drive ML research translating into product capabilities: prototyping, experiments, shipable features.
• Own models through full lifecycle: data strategy, training, evaluation, and integration into cloud CAD software Digital Surveyor.
• Develop evaluation metrics reflecting surveying and engineering accuracy requirements.
• Collaborate with ML infra engineers and product teams to scale training/inference and align with user needs.

Skills

Computer vision

Deep learning

Python

Model production

Communication

Education

Master's or PhD in ML/CS/related field

Tools

PyTorch

JAX

TensorFlow

Job description

The role

At Mach9, ML Engineers build the perception models at the core of our AI-enabled CAD system. We build models to extract 3D object and line features from dense LiDAR point clouds and imagery. Our unique data advantage allows us to develop and train cutting edge 3D scene understanding models that serve real surveyors and engineers in the field.

This role is both research-driven and product-focused. You'll design and train the models that power our automated extraction pipeline — image and 3D detection and localization — and work end-to-end from research prototype to production feature. You'll partner closely with infrastructure and product teams to take ideas from a paper to deployed capabilities.

This role is ideal for early-to-mid-career ML engineers who thrive on end-to-end ownership and are able to move fluidly from dissecting a new architecture paper to shipping the product feature that the resulting ML model backs.

Responsibilities

• Design, train, and evaluate computer vision and 3D ML models for extracting CAD-grade geometry and features from dense LiDAR and imagery.

• Drive ML research that translates directly into product capabilities: prototyping new approaches, running experiments, and identifying what’s shippable.

• Own models through the full product lifecycle: problem framing, data strategy, training, evaluation, and final integration into our cloud-based CAD software, Digital Surveyor.

• Develop evaluation methodology and metrics that reflect real surveying and engineering accuracy requirements.

• Work with ML infrastructure engineers to scale training and inference of your models and with product teams to align your model’s behavior with what the user wants.

Requirements

• Master's or PhD in Machine Learning, Computer Vision, Computer Science, or a related field, or equivalent industry experience.

• Strong foundation in computer vision and deep learning, with hands‑on experience training models for segmentation, detection, or 3D understanding.

• Experience taking a ML model from research/prototype to production, not just publishing or benchmarking.

• Working knowledge of geometric concepts relevant to 3D perception like coordinate systems and 3D transforms.

• Strong communication skills and the ability to collaborate with researchers, other engineers and product stakeholders.

• Proficient with Python and a production-quality ML library like PyTorch, JAX, or TensorFlow.

Bonus qualifications

• Experience with common 3D deep learning architectures, like point cloud backbones such as PTv3, sparse convolutions, or 3D detection/segmentation networks.

• Experience with large unstructured datasets — imagery and 3D point clouds — at scale.

• Experience delivering production-grade models with optimization techniques such as quantization, pruning, distillation, or runtime acceleration (e.g., TensorRT, ONNX Runtime).

• Familiarity with multi-GPU training and experiment management (Weights & WightBiases or similar).

• Publications or strong open-source contributions in computer vision or 3D machine learning.