Robotics Engineer, SLAM and Sensor Fusion Deployed on Real Robots

Pave Talent; Philadelphia (Philadelphia County); On-site; USD 140,000 - 190,000; Full time

Benefits offered by this job

Annual bonus

On-site in Philadelphia

Job summary

Pave Talent is hiring on behalf of our client, a venture-backed robotics company in Philadelphia, to design and deploy perception systems for real robots. You will own SLAM/localization, CV pipelines, and sensor fusion, with field data and production-grade software.

Candidates typically hold a master’s degree or PhD in robotics, CS, or EE with 2+ years of robotics experience, Linux, ROS/ROS 2, and hands-on work on real hardware.

Qualifications

• Master’s degree in robotics, CS, EE or related field with 2+ years of relevant experience, or PhD.
• Hands-on experience implementing SLAM or localization on a real system.
• Strong 3D geometry, coordinate transforms, camera models, features, and probabilistic state estimation.
• Proficiency in C++ and/or Python; live C++ exercise expected.
• Linux and ROS/ROS 2; robotics software stack.
• At least 2 years in a complex physical product domain, robotics, automation, or aerospace.
• Permanent U.S. residency and ability to clear background checks.
• On-site in Philadelphia five days a week; no remote option.
• SLAM frameworks: ORB-SLAM, VINS, Cartographer, or GTSAM.
• Lidar-based perception and mapping.
• Deep learning perception models.
• Perception deployed on real robots in the field.
• CUDA or OpenCL GPU acceleration.
• Dataset curation and annotation.
• Publications in robotics or computer vision with industry delivery.

Responsibilities

• Design and implement SLAM and localization systems, visual, visual-inertial, lidar, or multi-sensor
• Build computer vision pipelines for feature extraction, tracking, mapping, and scene understanding
• Implement and optimize estimation algorithms, both filtering and optimization-based
• Fuse data from cameras, inertial measurement units (IMUs), lidars, and depth sensors
• Evaluate perception performance against real-world field data and hard metrics
• Optimize for real-time performance and robustness on the robot's onboard compute
• Work with the controls and planning teams to support downstream autonomy
• Build out the tooling, datasets, and evaluation frameworks the team runs on

Skills

SLAM

Localization

C++

Python

Computer vision

ROS / ROS 2

Linux

GPU CUDA

Sensor fusion

Field deployment

Education

Master's degree in Robotics/CS/EE

PhD

Tools

Robot Operating System (ROS/ROS 2)

CUDA/OpenCL

Job description

One role, three levels. You do not pick yours before applying; the team places you at the level your skills and scope earn during the process, and the offer matches it.

Pave Talent is hiring on behalf of our client, a venture-backed robotics company under 200 people building autonomous ground robots for industrial inspection, infrastructure, energy, and public safety customers. The robots run on customer sites, on unstructured terrain, in real weather, today. Perception is the constraint on what the platform does next, which is why this hire matters more than the headcount suggests.

• Researchers who became engineers. You care about the algorithm and you care just as much about whether it runs at frame rate on the robot's compute
• People who ship. The team's hardest hiring problem is candidates who can explain a method beautifully and have never taken one to production
• Engineers who take direct feedback in a code review, change the approach, and move on
• Anyone who wants their work measured against physical terrain instead of a leaderboard

This is not the role if you want to spend two years deepening one research thread, or if your only deployment target has ever been an offline dataset.

You own a piece of the perception stack outright, not a ticket queue inside someone else's. Your algorithm goes on a robot within weeks and field data comes back. Lead is open now if you want direct reports. Senior and Staff stay fully individual contributor if you do not.

The platform is being pushed into harder environments and perception unlocks them. The team is expanding what the robot handles across perception and manipulation: graded stairs, grating, and bridge structures, the geometry that breaks most localization stacks. You take perception from research to reliable production software alongside the robotics, controls, and systems engineers.

• Design and implement SLAM and localization systems, visual, visual-inertial, lidar, or multi-sensor
• Build computer vision pipelines for feature extraction, tracking, mapping, and scene understanding
• Implement and optimize estimation algorithms, both filtering and optimization-based
• Fuse data from cameras, inertial measurement units (IMUs), lidars, and depth sensors
• Evaluate perception performance against real-world field data and hard metrics
• Optimize for real-time performance and robustness on the robot's onboard compute
• Work with the controls and planning teams to support downstream autonomy
• Build out the tooling, datasets, and evaluation frameworks the team runs on

This team has interviewed roughly 100 people across their open software roles and has not hired. The failures cluster: live C++ under observation, explaining a technical decision clearly, absorbing feedback in the room, and turning a concept into a finished product rather than an interesting prototype. The other consistent gap is candidates who cannot pivot off a long-horizon lab project into product cadence. Research background is welcome here; the hiring managers actively want lab experience. What they need with it is someone who can close.

• A master's in robotics, computer science, electrical engineering, or a related field plus 2+ years of relevant experience, or a PhD. Equivalent industry experience counts; the team weighs what you built and shipped over where your degree is from
• Hands-on experience implementing SLAM or localization on a real system, not just in simulation
• Solid command of 3D geometry and coordinate transformations, camera models and calibration, feature-based or direct visual methods, and probabilistic state estimation
• Proficiency in C++ and/or Python. Expect a live C++ exercise in the process
• Linux environments and a robotics software stack such as Robot Operating System (ROS / ROS 2)
• At least 2 years in a complex physical product domain, robotics, automation, or aerospace, or equivalent research lab work on real hardware
• Permanent U.S. residency, and the ability to clear standard background checks before and during employment
• On-site in Philadelphia five days a week. There is no remote option for this role
• SLAM frameworks: ORB-SLAM, VINS, Cartographer, or GTSAM
• Lidar-based perception and mapping
• Deep learning perception models
• Perception deployed on real robots in the field
• Graphics processing unit (GPU) acceleration with CUDA or OpenCL
• Dataset curation and annotation
• Publications in robotics or computer vision, paired with industry delivery experience

Five steps, and you will know where you stand at each one.

• A 30-minute technical screen with us at Pave Talent before anything goes to the client
• An initial phone screen with their people team
• A 45-minute technical conversation
• A live CoderPad coding interview with team members
• An on-site final focused on collaboration and fit

Annual Bonus: variable compensation that has averaged at least 15% of base, prorated in year one. This is the actual historical average, not a stretch target

Travel: none required, though domestic and international field opportunities exist if you want them