Computer Vision Engineer

Join the research team at Tutor Intelligence to build perception systems that work on real robots. You will combine

geometric methods and learned visual models to turn camera and sensor data into reliable inputs for robot

manipulation, experimentation and evaluation.

The role connects algorithms with physical systems. You will work across camera setup and calibration, data quality,

model development, evaluation and deployment, with responsibility for understanding failures in the actual robot

environment.

Requirements

Develop and integrate perception capabilities for robot manipulation, such as detection, segmentation, tracking

and 3D scene understanding.

• Calibrate cameras and sensor relationships; work with coordinate frames, synchronization and measurement

uncertainty.

• Build data and evaluation pipelines that expose failure modes across changes in objects, lighting, viewpoint and

occlusion.

• Adapt learned models and geometric methods to the task, selecting approaches based on measured performance

and practical constraints.

• Deploy perception into robot software, balancing accuracy, latency, robustness and maintainability.
• Debug failures with mechanical, electrical, software and research colleagues, and make experiments reproducible.

Strong Python programming and experience building computer-vision systems beyond an isolated notebook or

benchmark.

• Working knowledge of camera models, calibration, coordinate transforms and 3D geometry.
• Experience with geometric vision, learned visual models, or both, and the ability to evaluate tradeoffs rigorously.
• Familiarity with tools such as OpenCV, Open3D or PCL and a modern deep-learning framework where applicable.
• Ability to build meaningful datasets and evaluations, diagnose errors and explain improvements with evidence.
• Comfort connecting perception to robotics concepts such as kinematics, robot models, planning and control.

Nice to haves (zero or more)

Deployment on a physical robot or another system using live sensor data.

Multi-camera systems, depth sensing, sensor fusion, SLAM, object pose estimation or manipulation.

C++, GPU optimization or efficient deployment of visual foundation models.

What success looks like:- Establish a reproducible baseline and evaluation for a concrete perception problem.

• Deliver an improvement that survives integration and testing on the physical system.
• Make failure modes, calibration assumptions and latency constraints visible to the rest of the team.

About our Roles & Titles

At Tutor, we believe great engineers and researchers are defined by what they build and the impact they have — not where they sit in an org chart or what title they have. Therefore, everyone in our R&D org holds the title Member of Technical Staff (MoTS). Our job postings use standard titles so you can find us, but if you join Tutor, you'll be a MoTS — with a level that is determined through the interview process.

That also means we hire people, not slots. Work at Tutor evolves every quarter, and we set the expectation of flexibility from day one — it's common for people to start on one thing and shift to another based on where the team needs them most. A high technical bar across the board is what makes that flexibility possible: it's what allows people to contribute meaningfully whatever problem they take on.