Senior Machine Learning Engineer: Computer Vision & 3D

FitMatch; Fort Lauderdale (FL); Remote; USD 160,000 - 200,000; Full time

Job summary

FitMatch seeks a Senior Machine Learning Engineer to advance QuadraScan's computer vision and 3D reconstruction. You will own model development end to end, from data prep to production, on a small team reporting to the Head of Mobile Development.

This remote, IC role emphasizes deep hands-on work, evidence-based decisions, and collaboration with mobile and backend engineers to balance accuracy, latency, and cost.

Qualifications

• 5+ years of professional ML/vision experience with hands-on model development.
• Proven track record deploying CV models to production based on real-world data.
• Strong Python and PyTorch skills for building training pipelines.
• Experience with geometric CV (depth estimation, camera calibration, 3D reconstruction).
• Solid software engineering fundamentals and ability to work independently.

Responsibilities

• Own model development for human reconstruction from data preparation to production integration.
• Select and fine-tune pretrained models, adjusting architectures when needed.
• Design experiments with clear baselines and success criteria.
• Improve training data and maintain dataset versioning.
• Diagnose failures across the pipeline from capture quality to model predictions.
• Evaluate geometry fidelity and robustness.
• Build maintainable training and inference code with reproducible experiments.
• Collaborate with mobile and backend engineers to deploy improvements.
• Communicate findings with clear evidence and recommendations.

Skills

Python

PyTorch

Computer Vision

3D Reconstruction

Data-driven decision making

Independent work

Job description

Role Overview

FitMatch is seeking a Senior Machine Learning Engineer to advance the computer vision and 3D reconstruction behind QuadraScan, starting with improving how we reconstruct human geometry from ordinary camera images, focusing on geometric accuracy, reliable measurements, and real-world performance. You'll own model development end to end, from data preparation and experimentation through evaluation and shipping, building on existing prototypes, datasets, and tools with real influence over our methods.

This is a hands-on senior IC role on a small team, reporting to the Head of Mobile Development. You'll align with them on objectives and trade-offs while independently driving ML implementation and technical decisions. We expect strong, evidence-backed recommendations and a track record of seeing work through to delivery.

Location:

Remote

Base salary:

$160,000–$200,000, depending on experience and work location

Responsibilities:

• Own model development for human reconstruction and related computer vision work, from data preparation and training through evaluation and production integration

• Select, adapt, and fine-tune pretrained models, modifying architectures or training objectives when a clear product need justifies it

• Design focused experiments with clear baselines, success criteria, and sensible time and compute limits

• Improve training data and capture coverage across synthetic and real imagery, maintaining reliable labels, dataset versioning, and train/eval subject separation

• Diagnose failures across the full pipeline, from capture quality and camera geometry to model predictions and reconstructed surfaces

• Evaluate geometric fidelity, regional errors, robustness, and failure rates, tying model metrics to scan and measurement quality

• Build maintainable training and inference code with reproducible experiments and practical regression checks

• Partner with mobile and backend engineers to deploy improvements, balancing accuracy, latency, memory, and cost

• Communicate findings clearly, including what the evidence supports, what remains uncertain, and the best next step

Qualifications:

• 5+ years of professional experience in ML, computer vision, or related engineering, with substantial hands-on model development

• Proven track record taking a computer vision model from experimentation into production and improving it based on real-world performance and failure cases

• Strong Python and PyTorch skills, including building training pipelines, modifying model components and losses, and diagnosing training issues

• Practical experience with geometric computer vision (e.g., depth estimation, camera calibration, multi-view geometry, 3D perception, or reconstruction)

• Solid grasp of model evaluation, including overfitting, data leakage, distribution shift, and the limits of aggregate metrics

• Strong software engineering fundamentals: maintainable code, version control, testing, debugging, and reproducible environments

• Ability to turn ambiguous objectives into actionable technical plans and execute independently

• Clear communication and sound judgment in balancing model quality, speed, and resources

Preferred Qualifications:

• Experience with human reconstruction, body shape, pose estimation, photogrammetry, or measurement-sensitive vision applications, including work with point clouds, meshes, or differentiable rendering

• Experience using synthetic data to improve transfer to real camera imagery, and adapting pretrained vision transformers or other large vision models to specialized tasks

• Familiarity with deploying models on Apple platforms (Core ML, MLX, Metal) or other constrained devices

• Experience delivering applied ML in a small company or a team with limited specialist support

How We Work:

• We organize work around concrete product improvements, using experiments to drive decisions and following promising results through to reliable implementation.

• Technical direction is collaborative. We value engineers who explain their reasoning, flag problems early, and push for better approaches when the evidence supports it.

• We keep tooling proportionate to the work. Clear evaluation, reproducible results, and dependable delivery matter more than the size of the stack.

• AI-assisted engineering is standard at FitMatch. Tools like Claude Code are part of our daily workflow, and we expect engineers to use them thoughtfully while owning the correctness of the resulting code and conclusions.