AI Engineer

Cyvl•Somerville, MA

7d•$125,000 - $225,000•Hybrid

About The Position

Cyvl is the Physical AI company changing how infrastructure is managed. We build purpose-built sensors, computer vision and AI that turn every drive into current data: what exists, where it is and what condition it's in. Our Infrastructure Intelligence Platform is used by over 500 cities and towns to direct work and fix problems before they fail. As an AI Engineer, you'll turn Cyvl's large proprietary datasets of US infrastructure (LiDAR and imagery from hundreds of cities) into intelligence. This includes developing computer vision models for asset detection and assessment, as well as LLM-powered agents and tools for customers and partners to query the physical world. You will be involved in the entire lifecycle of models, from prototype to production, shipping AI features that customers use daily. This role is based in Somerville, MA, or remote from San Francisco, CA.

Requirements

3+ years building and shipping ML or AI systems in production
Strong Python and a modern deep learning framework (PyTorch preferred)
Hands-on experience with computer vision, 3D perception, or LLM applications (RAG, tool use, agents, evals)
Solid software engineering fundamentals: testing, code review, and cloud deployment
Ability to go from an ambiguous problem to a working prototype quickly, then make it production-grade

Nice To Haves

Point cloud or multimodal work (LiDAR and imagery fusion)
Model Context Protocol (MCP), foundation model APIs, or agent frameworks
Geospatial or remote sensing background
MLOps: model serving, GPU infrastructure, experiment tracking
Advanced degree in CS, ML, robotics, or a related field, or equivalent experience

Responsibilities

Build and improve computer vision and 3D perception models that detect, classify, and assess infrastructure (pavement, signs, sidewalks, trees) from LiDAR and imagery.
Build LLM-powered products: agentic workflows, our MCP server, and AI copilots that put Cyvl data to work inside customer tools.
Own models from data and training through evaluation, deployment, and monitoring in production.
Design evaluation sets and metrics that tie model performance to what customers actually need.
Stay close to the state of the art and bring the right new tools into our products and our own workflows.