Research Robotics

We're Hiring: Research Engineer – Computer Vision & Egocentric Perception | Physical

AI

We are looking for a Research Engineer specializing in Computer Vision and Egocentric Perception to join an innovative AI organization developing intelligent systems that understand human actions and real-world environments.

The organization operates at the intersection of embodied AI, robotics, computer vision, and multimodal data, building advanced perception technologies for real-world applications.

Why Join

Work on challenging perception problems that go beyond conventional computer vision benchmarks. You'll have the opportunity to build proprietary models, work with real-world egocentric video data, and take research innovations from experimentation to production.

This role offers ownership across model architecture, training data, evaluation, deployment, and performance optimization.

Location: Bangalore

Experience: 4+ Years

Key Responsibilities:

• Develop and train custom computer vision models for 3D hand pose, human pose estimation, depth estimation, and visual perception.
• Improve model performance on challenging real-world video data involving occlusion, motion blur, wide-field cameras, and complex environments.
• Build and improve perception pipelines for pose estimation, depth reconstruction, and object detection.
• Develop reliable ground-truth and model evaluation methodologies.
• Work on stereo vision, camera calibration, multi-view geometry, and visual localization.
• Build and optimize training and inference pipelines using AWS and GPU infrastructure.
• Develop automated quality assessment models to identify errors in AI-generated annotations.
• Take ownership of model development from data curation and experimentation through production deployment.

Must-Have Skills:

• 4+ years of experience in Computer Vision, Robotics, or Machine Learning.
• MS/PhD in Computer Vision, Robotics, or ML, or equivalent published research experience.
• Strong hands-on expertise in Python and PyTorch.
• Experience training and deploying custom perception models that outperform existing baseline models on real-world data.
• Strong understanding of multi-view geometry, camera calibration, triangulation, and coordinate transformations.
• Production inference experience on AWS, including containerization and batch processing.

Deep expertise in at least two of the following:

• 3D Hand/Human Pose Estimation
• Stereo or Monocular Depth Estimation
• Visual-Inertial Odometry (VIO) / SLAM
• Object Detection and Segmentation
• Video Vision-Language Models (VLMs)

Preferred Skills:

• Research publications in egocentric vision, 3D hand pose, VLM/VLA models, or robot learning.
• Experience with IMU-synchronized multimodal data and camera-IMU calibration.
• Familiarity with robotics data formats such as MCAP, Foxglove, ROS, LeRobot, or RLDS.
• Experience working with datasets such as Ego4D, EgoExo4D, or Open X-Embodiment.
• Open-source contributions in computer vision or robotics.

Ideal Candidate:

A research-oriented engineer who enjoys solving complex perception challenges, has independently developed and deployed computer vision models, and is comfortable taking ownership of the complete model lifecycle - from research and experimentation to production

delivery.

If you're passionate about Computer Vision, Robotics, and building AI that understands the physical world, we'd love to connect. Apply now or share your updated resume to [Confidential Information]