Software Engineer - Onboard Maps

Applied Intuition•Sunnyvale, CA

1d•$170,000 - $265,000•Onsite

About The Position

Applied Intuition, Inc. is powering the future of physical AI. Founded in 2017 and now valued at $15 billion, the Silicon Valley company is creating the digital infrastructure needed to bring intelligence to every moving machine on the planet. Applied Intuition services the automotive, defense, trucking, construction, mining and agriculture industries in three core areas: tools and infrastructure, operating systems, and autonomy. Eighteen of the top 20 global automakers, as well as the United States military and its allies, trust the company’s solutions to deliver physical intelligence. Applied Intuition is headquartered in Sunnyvale, California, with offices in Washington, D.C.; San Diego; Ft. Walton Beach, Florida; Ann Arbor, Michigan; London; Stuttgart; Munich; Stockholm; Bangalore; Seoul; and Tokyo. Learn more at applied.co. We are an in-office company, and our expectation is that full-time employees primarily work from their Applied Intuition office 5 days a week. However, we also recognize the importance of flexibility and trust our employees to manage their schedules responsibly. This may include occasional remote work, starting the day with morning meetings from home before heading to the office, or leaving earlier when needed to accommodate family commitments. This in-office expectation does not apply to contractor positions. About the role Our Localization, Calibration, and Mapping team delivers high-integrity solutions that let autonomous systems navigate with precision in real-time, safety-critical environments — across consumer automotive, trucking, mining, and construction. As a new team focused on platformization, we offer a rare opportunity to own core work leveraged across the entire organization. HD maps go stale the moment they're built — lanes get repainted, stop lines move, roads close. You'll own the live map on Applied's onboard stack: a real-time fused estimate of the world built from map priors and live perception, from the estimation core to the change-detection and trust semantics the planner consumes, across every vehicle vertical we serve. This layer is early and has no incumbent design. You'll set the architecture and define the interfaces rather than tune inside someone else's system. We encourage engineers to take ownership of technical and product decisions and to interact closely with users to collect feedback.

Requirements

5+ years building production software, including at least one shipped system where they owned the estimation core
State estimation fundamentals proven in production: factor graphs and/or filtering, data association, uncertainty representation and propagation, and 3D geometry and transforms
Experience fusing heterogeneous sources such as prior map data, perception output, and raw sensors, into a single world estimate, with rigorous error characterization
Strong modern C++ and the ability to make estimation code meet real-time budgets on embedded compute
Rigorous evaluation practice: has built ground-truth or metrics pipelines and used them to hold a shipped operating point and prevent regressions
Judgment about uncertainty: understands the difference between an error that degrades gracefully and one that is catastrophic, and designs the system accordingly
Experience defining a system and its interfaces and negotiating those contracts with consumer teams

Nice To Haves

Map change detection or live/lifelong mapping experience
HD map semantics: lane graphs, topology, map formats and their failure modes
GTSAM or comparable factor-graph frameworks; Lie group methods
Degraded-mode or integrity-monitor design for a safety-relevant function
MS/PhD with a focus on state estimation or SLAM

Responsibilities

Own the onboard live-map estimator end to end: fuse HD map priors with live perception into a single, real-time estimate of the world the rest of the stack can trust
Build our onboard map change detection: classify situations like boundary shifts, stop-line moves, and road closures from the divergence between map and perception, and define when the planner should stop trusting the map
Define the map trust contracts the planner builds against: confidence semantics, degraded-mode behavior, and the operating points that hold across vehicle programs
Build the evaluation backbone, ground-truth and metrics pipelines, operating points, and regression testing that keep estimator and detector quality from silently drifting
Ship estimation code in modern C++ that meets real-time budgets on embedded compute
Extend the loop offboard over time: aggregate divergence detections across fleet drive logs into map freshness and correction pipelines