SLAM Engineer vs Perception Engineer: Related specialisms, different searches

Object detection, tracking, classification, semantic segmentation, sensor fusion, and depth estimation. Their output is "what is in the scene and where is it relative to the robot."

Perception engineers typically work with camera and LiDAR data, often fusing multiple sensor modalities to produce a coherent understanding of the environment. They build the systems that allow a robot to detect obstacles, identify objects of interest, and understand the semantic content of a scene.

The work is heavily ML-dependent. Most modern perception pipelines use deep learning for detection and segmentation, with classical methods for fusion and tracking. Production perception engineers need both ML depth and systems engineering rigor to deploy models that run reliably in real time on constrained hardware.

Both work with LiDAR and camera data. Both need strong C++ and mathematics. Both deal with noisy real-world sensor data. Both must produce outputs that are reliable enough to feed into downstream systems that make safety-critical decisions.

In smaller teams, one person might do both. This is common in early-stage startups where a single engineer owns the entire sensing and localization stack. In larger teams, they are distinct roles with different reporting lines and different interview processes.

Visual SLAM in particular overlaps with perception. Feature extraction, image matching, and depth estimation are used in both disciplines. An engineer working on visual odometry may share tools and techniques with one working on object detection, even though the downstream objectives are different.

SLAM engineers are rarer. The specialism is narrower and the academic pipeline is smaller. A handful of top research groups globally produce the majority of strong SLAM practitioners: CMU in Pittsburgh, ETH Zurich, Oxford, MIT, and a few others.

Perception has more crossover from general computer vision and ML. Engineers from large technology companies or ADAS programs can often transition into robotics perception roles with moderate ramp-up. This makes the effective talent pool for perception significantly larger.

In practice, this means SLAM searches typically take longer and require deeper mapping of the autonomy and navigation market. The candidates are fewer, more concentrated in specific institutions and companies, and less likely to be visible on job boards or LinkedIn. A specialist search approach is almost always necessary for SLAM, whereas some perception roles can be filled through broader channels.

If your robot needs to understand what is around it (detect objects, avoid obstacles, identify targets), you need perception. If your robot needs to know where it is and build a map of its environment, you need SLAM. If your robot needs both, you need both, and they will work closely together.

For early-stage companies with a small team, hiring a strong robotics software engineer who can handle both at a basic level may be pragmatic. But as the system matures, dedicated perception and SLAM engineers will be necessary. Trying to keep one person across both at production quality is not sustainable.

When writing the job brief, be specific about which one you actually need. A brief that asks for "perception and SLAM" as a combined role will attract neither the best perception engineers nor the best SLAM engineers. Both groups will read it as a company that does not understand the distinction.