AI Multimodal Scene Understanding Engineer

Prixel Labs

Remote · $160k – $200k / year · Full-time · Posted 4 months ago

This role is no longer accepting applications

Don't miss the next role like this

Create a free profile with your skills. We match you to new AI roles as they open, and companies hiring can contact you directly.

Create free profile, 2 minutes

Similar open roles you can apply to now:

See all open AI jobs

This role is no longer accepting applications

Don't miss the next role like this

Create a free profile with your skills. We match you to new AI roles as they open, and companies hiring can contact you directly.

Create free profile, 2 minutes

Similar open roles you can apply to now:

See all open AI jobs

About the role

What does it mean to truly understand a scene? Our platform processes 3D scans, RGB images, and sparse depth maps from 60 enterprise clients in architecture, construction, and insurance. A human expert looking at a building exterior can identify material types, assess structural integrity, spot water damage, and estimate renovation cost — all from visual information. We want our system to do the same. Scene understanding at this level requires fusing multiple sensing modalities with learned structural priors. You'll build the ML stack that takes raw multimodal inputs and produces semantic scene graphs: labeled objects with spatial relationships, material classifications with confidence estimates, and anomaly flags for the features that human reviewers need to examine more closely. The research team has laid the architectural groundwork — you'll take their prototypes and turn them into production systems that process 4,000 scans per day with SLA-bound latency and documented accuracy. This is a senior IC role where your technical depth in multimodal perception directly determines what the product can offer.

Fully remote.

Responsibilities

Build production-grade multimodal scene understanding models from 3D and RGB inputs
Develop semantic scene graph generation with object relationships and material labels
Implement anomaly detection for structural and material assessment use cases
Optimize multi-modal inference to meet SLA requirements at 4,000 scans/day
Build evaluation pipelines benchmarking against human expert annotations
Collaborate with research to bring prototype architectures to production readiness

Requirements

6+ years in computer vision or multimodal ML with production deployment experience
Deep expertise in 3D vision: point clouds, depth estimation, or NeRF-based methods
Experience fusing heterogeneous sensor modalities in production systems
Strong PyTorch skills and understanding of attention-based architectures
Track record shipping models that run at scale with documented accuracy