Machine Learning / Computer Vision Engineer
Full Time
|
Cambridge, MA
|
Eka Robotics
Eka Robotics is on a mission to build intelligence for the physical world by creating robots that are fast, general-purpose, and reliable. Our physics-based approach enables robots to achieve superhuman capabilities and pushes the frontier of robotics research and deployment.
Our team is composed of pioneers in robotics and machine learning. As we continue to scale our research and development efforts, we're looking for hands-on engineers who are excited to help shape the future of robotics.
Responsibilities
Build computer vision and visual representation learning pipelines for robotic manipulation using:
RGB and RGB-D imagery
Depth sensing
Semantic segmentation
Pose estimation
Keypoint detection
Object-centric representations
Develop visual models that support reinforcement learning and imitation learning, including end-to-end visuomotor policies that map visual observations directly to robot actions.
Improve vision data pipelines through:
Domain randomization
Photorealistic rendering
Synthetic data generation
Sensor noise modeling
Real-world fine-tuning
Design and train perception models that are robust to:
Lighting variations
Camera viewpoint changes
Texture variation
Clutter and occlusion
Object instance diversity
Imperfect sensor calibration
Evaluate learned visual representations and manipulation policies on real robotic systems, identify failure modes, and iterate on models, datasets, and training strategies.
Collaborate closely with robotics, robot learning, and simulation engineers to define perception strategies for robotic manipulation.
Configure, calibrate, and evaluate camera and depth sensing systems, with an emphasis on how sensor selection impacts learned policies and real-world robustness.
Minimum Qualifications
Ph.D. in Computer Vision, or 3+ years of experience developing production computer vision systems.
Strong background in machine learning for computer vision, particularly deep learning-based visual perception.
Experience training modern computer vision models using frameworks such as:
JAX
PyTorch
or equivalent frameworks
Practical experience with one or more of the following:
Visual representation learning
Object detection
Semantic segmentation
Pose estimation
Depth estimation
Object tracking
3D perception
Strong Python programming skills.
Ability to transition seamlessly between research prototypes and production-quality software.
Strong understanding of how factors such as data distribution, sensor noise, calibration, lighting, and scene variation influence model performance.
Preferred Qualifications
Experience with one or more of the following:
Training robot policies from visual observations, including:
RGB
RGB-D
Point clouds
Object-centric representations
Learned latent representations
Domain randomization, synthetic data generation, differentiable rendering, neural rendering, or photorealistic simulation.
Robotics simulation platforms such as:
Isaac Sim
MuJoCo
or similar environments
Robot learning techniques including:
Reinforcement learning
Behavior cloning
Diffusion policies
Offline reinforcement learning
Learning from demonstrations
Deploying perception systems on real robots, including:
Camera calibration
Hand-eye calibration
Depth sensors
ROS or ROS2
Robot data collection pipelines
First-author publications at leading computer vision, robotics, or machine learning conferences, including:
CVPR
ICCV
ECCV
NeurIPS
ICLR
ICML
RSS
CoRL
ICRA
IROS