Computer Vision Engineer

Panoculon Labs Pvt. Ltd.; India; On-site; INR 1,500,000 - 2,700,000; Full time

Job summary

Panoculon Labs Pvt. Ltd. is seeking an experienced OpenCV PyTorch SLAM/VIO engineer to design perception algorithms that turn multi-sensor data into calibrated geometry, depth and scene understanding for robotics applications.

You will own vision problems end-to-end from projective geometry to on-device models. The role emphasizes strong CV fundamentals, real-time processing, and deployment of ML models on-device using TensorRT, PyTorch, and OpenCV.

Qualifications

• Solid computer-vision fundamentals: epipolar and stereo geometry, homography, rectification, intrinsic/extrinsic calibration.
• Strong math: projective geometry, linear algebra, and optimization.
• Image signal processing to extract meaningful information from raw data.
• Familiar with camera concepts (AWB, exposure, lenses).
• Proficient in Python and C++, with OpenCV.
• ML / deep learning experience (PyTorch): training, evaluating, and deploying models.
• 3D vision: point clouds, LiDAR, SLAM, or visual-inertial odometry.
• Robotics learning / VLA models and egocentric vision.

Responsibilities

• Camera calibration, stereo depth, epipolar geometry, rectification, and undistortion across single- and multi-camera systems.
• Homography and reference-frame transformations.
• Segmentation, detection, and tracking pipelines that blend classical CV with modern deep learning.
• Deploying models on-device - quantization and edge/mobile inference - and fusing camera, IMU, and LiDAR streams into the application.
• VIO / SLAM.

Skills

Python

C++

OpenCV

PyTorch

SLAM

VIO

Computer Vision

Linear Algebra

Projective Geometry

TensorRT

Job description

OpenCV PyTorch SLAM / VIO C++ & Python On-device ML

You design and build the perception algorithms at the heart of our stack - turning raw multi-camera, LiDAR, and IMU data into calibrated geometry, depth, and scene understanding that robots and wearables can act on. You'll own vision problems end to end, from the underlying projective geometry to models running on-device.

What you'll get to work on

• Camera calibration, stereo depth, epipolar geometry, rectification, and undistortion across single- and multi-camera systems.
• Homography and reference-frame transformations.
• Segmentation, detection, and tracking pipelines that blend classical CV (filtering, features, denoising) with modern deep learning.
• Deploying models on-device - quantization and edge/mobile inference (TensorRT, CoreML, TFLite) - and fusing camera, IMU, and LiDAR streams into the application.
• VIO / SLAM.

What we're looking for

• Solid computer-vision fundamentals: epipolar and stereo geometry, homography, rectification, intrinsic/extrinsic calibration.
• A solid grip on the math underneath - projective geometry, linear algebra, and optimization.
• Image signal processing - altering raw image data to get useful information.
• Knows cameras (AWB, exposure, how lenses work).
• Strong coding in Python and C++, and fluency with OpenCV.
• ML / deep learning experience (PyTorch): training, evaluating, and deploying models.
• 3D vision - point clouds, LiDAR, SLAM, or visual-inertial odometry.
• Robotics learning / VLA models and egocentric vision.