Robotics Engineer - Autonomy & Hardware Integration

Scion Ventures; Fremont (CA); On-site; USD 120,000 - 180,000; Full time

Job summary

Scion Ventures is seeking an engineer to bridge hardware and autonomy for an autonomous robot operating in real commercial indoor environments.

You will own sensor mounting, calibration routines, onboard ROS 2 autonomy stack, and field-grade debugging to ensure robust operation on real hardware.

Qualifications

• 4–7 years building robots that operate outside a lab.
• Deep ROS 2 knowledge: navigation, transforms, frames, cost maps, sensors, builds.
• Hands-on mechanical work: CAD, 3D printing, fasteners, tolerances.
• LiDAR-inertial SLAM and localization in production environments.
• Manipulation experience: grasp planning and arm kinematics on real hardware.
• Camera and LiDAR calibration, both offline and online routines.
• Production-grade motion control and fault diagnosis in the field.
• Field debugging under pressure to isolate faults to the correct layer.

Responsibilities

• Sensor integration and mechanical rig design that survives transport and vibration.
• Calibration tooling and runbooks to make recalibration routine.
• Onboard autonomy stack integration with ROS 2 across subsystems.
• Navigation behavior: multi-goal execution and obstacle handling.
• Manipulation on mobile platform with accurate base positioning.
• Field operation: run trials, collect data, diagnose on-site.
• Manage power and thermal budgets as a testing constraint.

Skills

ROS 2 proficiency

Mechanical competence

LiDAR-inertial SLAM

Manipulation & grasping

Camera/LiDAR calibration

Motion control tuning

Field debugging

Tools

CAD

3D printing

ROS 2 tooling

Job description

Scion Ventures is hiring on behalf of one of our portfolio companies, an early-stage robotics team building autonomous systems that operate in real commercial indoor environments. Not simulation, not a lab. There is hardware on the ground today.

The company is not named publicly at this stage. We will tell you who it is on the first call.

THE ROLE

They are hiring one engineer to own the seam between hardware and autonomy.

Most robotics teams split these functions across two or three people. At this company's size, that split is the problem: mechanical issues get diagnosed as software bugs, calibration drift gets diagnosed as an algorithm failure, and days disappear into the gap. They want one person who can stand in front of the robot, tell which layer is actually broken, and fix it themselves.

WHAT YOU WILL OWN

• Sensor integration and mechanical rigidity. Mounting design that survives transport, vibration, and a moving platform. Mounting geometry computed against detection-range and field-of-view requirements rather than estimated.
• Calibration. The extrinsic calibration procedure, the tooling, and the runbook, so recalibration is a routine operation instead of an outage.
• Onboard autonomy stack. ROS 2 integration across navigation, localization, perception, and low-level control. Runtime and build reliability, including containerized environments.
• Navigation behavior. Multi-goal execution, state management between goals, goal-pose orientation accuracy, obstacle handling, and coordination between navigation and other onboard subsystems.
• Manipulation. Grasping and arm control on a mobile platform, including the base-positioning accuracy that grasping depends on.
• Field operation. Be physically with the hardware. Run trials, capture data, diagnose on the spot, and turn field failures into fixes rather than tickets.
• Power and thermal budget as a first-class constraint on what can be tested.

WHAT THEY NEED

Required:

• 4 to 7 years building robots that operate outside a lab
• Deep ROS 2: navigation, transforms and frames, cost maps, sensor drivers, launch and build systems. You can debug a broken transform tree without help
• Hands-on mechanical competence: CAD, 3D printing, fastener and tolerance selection, vibration and rigidity. No ME degree required, but your brackets hold
• LiDAR-inertial SLAM and localization in production. You understand the failure modes when point-cloud processing exceeds the real-time budget
• Manipulation experience: grasp planning, arm kinematics, gripper control on real hardware
• Comfortable being the only person in the building when something breaks
• Camera and LiDAR calibration, offline and online. Intrinsics and extrinsics on the bench, plus online routines that catch drift during operation and re-estimate without pulling the robot out of service
• Motion control in production. You have tuned path tracking on real hardware with pure pursuit, DWB, or Stanley, and can diagnose tracking error in the aisle rather than in simulation
• Field debugging under pressure. You isolate a fault to the right layer, mechanical, sensor, driver, estimator, planner, or controller, with the robot stopped in front of you and a customer waiting

Strong bonus:

• Commercial mobile-manipulation or legged platform experience
• Logistics, inventory, or facility robotics
• Visual SLAM, or camera and range-sensor fusion for close-range perception
• Bipedal or quadrupedal platform bring-up, including mapping and locomotion on real hardware
• Gait tuning on legged robots: stability, footstep placement, disturbance recovery
• Dexterous bimanual manipulation. Two arms or hands coordinated to grasp, reorient, and place objects

Not required:

• A PhD
• Simulation-only experience

FIRST 90 DAYS

• Day 30: sensor mounting and calibration solved and documented. Localization holds through a full operational loop without operator intervention.
• Day 60: multi-goal navigation runs end to end reliably, correct final orientation, onboard perception integrated with the rest of the system.
• Day 90: the robot performs a full physical task on demand, in the real environment. You can say honestly what the platform will and will not do next.

1. Screen, 30 min. Background, plus one question: tell us about a robot that failed in the field and what the root cause turned out to be.

2. Technical deep dive, 60 min. We describe a real failure. You debug it out loud. We are watching whether you reason from the system or pattern-match from memory.

3. Hardware judgement, 45 min. A real mounting and geometry problem. Bring the tradeoffs, not just the answer.

4. On-site half day with the actual hardware. As much for you as for them.

5. Two reference calls.

No take-home. No algorithm puzzles.

HONEST NOTES

• Early-stage company. Some weeks are CAD and fasteners. Some weeks are C++.
• You will not own the cloud or backend platform. You integrate with that boundary.
• They work with an outside specialist on parts of the autonomy stack. Absorbing that knowledge into the team is part of your first quarter.

Location: On-site, Fremont, CA. Full-time. Compensation: base plus meaningful early equity.