Machine Learning Engineer - Multimodal Modeling

Stand Insurance•San Francisco, CA

3d•$250,000 - $295,000•Onsite

About The Position

As a Machine Learning Engineer on the Applied Science team, you will design, train, and deploy Stand's flagship AI capabilities, with a central focus on the multimodal meshing of our Stand World Model with powerful language models. This work brings physical simulation, rich 3D representations of real assets, and broader business context together into models that can reason across all of them at once, in support of better underwriting, pricing, and mitigation decisions. This is a hands-on, high-ownership position on the Machine Learning team within Stand Applied Science. You will own modeling work end-to-end, from architecture and training strategy through evaluation and production deployment, and partner closely with the Platform team to ensure the agentic harness and workflows your models plug into deliver strong results in production. Your partnerships will extend across the business, mirroring the breadth of the model's inputs: collecting technical insight from subject matter experts and other MLEs, and institutional judgment from underwriting, pricing, mitigation, inspection, and customer decision-making.

Requirements

Deep hands-on experience designing and training multimodal models, fusing heterogeneous data (e.g., 3D/vision, simulation outputs, tabular, and text) into shared representations
A record of bringing models of this class to production: training at scale, evaluation, deployment, and iteration on live systems
Experience applying ML to complex physical systems. We are agnostic to the domain: atmospheric, molecular, protein, robotics, fluid dynamics, or other physics-grounded modeling all carries over
Experience training or fine-tuning LLMs, including tool use, agentic workflows, or post-training methods
Strong project ownership and execution: planning, prioritization, and delivery of complex technical work
Ability to operate across disciplines, connecting technical development to business objectives
Strong, succinct communication and judgment to balance R&D, delivery timelines, and business impact
Highly self-motivated, proactive, and adaptable; comfortable in fast-paced, ambiguous environments
Candidates must be authorized to work in the U.S. Stand does not sponsor new work visas. We can consider candidates on TN visas, O-1A visas, or H-1B transfers with three years or more remaining.

Nice To Haves

Experience with retrieval and embedding systems: vector search and similarity in latent space
Experience with geometric deep learning: point clouds, meshes, and spatially-aware architectures
Familiarity with physics-informed AI and surrogate modeling across a number of domains
Experience in startups or zero-to-one technology development
Knowledge of geospatial, remote sensing, or Earth observation datasets

Responsibilities

Designing and training multimodal model architectures that jointly reason over physical, spatial, and business-context data
Building frameworks that let these models act as agents within nuanced workflows, making complex tool calls that include interacting with our world-modeling stack
Developing retrieval and similarity capabilities over learned representations of real-world assets and their multi-layered complexities
Standing up the training-data pipelines, evaluation harnesses, and production monitoring that take these models from prototype to production, collaborating closely with Applied Science infrastructure engineers to build out proper tooling
Design, build, and deploy machine learning systems spanning multimodal learning, physics-informed AI, digital twins, and spatial intelligence, contributing directly to core business impact
Own projects end-to-end, from problem definition and prototyping through production deployment, adoption, and ongoing performance monitoring
Develop rigorous evaluation frameworks that weigh model judgments against real business outcomes
Build on and extend scalable ML infrastructure
Partner with Stand’s Platform team on the model-harness interface
Drive cross-functional alignment, communicating decisions, tradeoffs, and status