Computer Vision Engineer

Cosmoserve Space; Hyderabad; On-site; INR 1,500,000 - 2,600,000; Full time

Job summary

Cosmoserve Space invites a Computer Vision Engineer to develop perception systems for spacecraft, estimating 6-DoF poses of space objects using classical vision and deep learning in simulation environments.

You will build pipelines for object detection, pose estimation, and domain randomization, and deploy models on NVIDIA Jetson with TensorRT for real-time inference, collaborating with GNC and simulation teams.

Qualifications

• Master’s degree or higher in Computer Vision, Robotics, or CS.
• 2+ years (or 4+ years) of relevant experience in CV/robotics with strong ML background.
• Strong foundation in computer vision and deep learning, CNN backbones and pose estimation.
• Proficient in Python, PyTorch, and OpenCV; familiar with data augmentation and domain randomization.

Responsibilities

• Build vision-based perception pipelines for pose estimation and object detection.
• Develop and deploy models for stereo/monocular camera geometry and depth estimation.
• Work with PnP, RANSAC, and geometry to recover 6-DoF poses of space objects.
• Optimize and deploy models on embedded platforms (NVIDIA Jetson) with TensorRT.

Skills

Computer Vision

Deep Learning

Python

PyTorch

OpenCV

CUDA

Jetson

TensorRT

Blender

Domain Randomization

Education

Master’s degree in Computer Vision / Robotics / CS

Tools

NVIDIA Jetson

TensorRT

ONNX

CUDA

Blender

Python

Job description

Cosmoserve Space is dedicated to building a cleaner, more sustainable future for space exploration through advanced active debris removal technologies. The company’s mission is to safeguard the orbital environment with high-tech, cost-effective solutions that address the growing challenge of space debris. Cosmoserve Space combines cutting-edge engineering with a market-leading approach to responsible space operations. Team members collaborate in a passionate, expert-driven environment that values innovation, reliability, and long-term impact. The organization is focused on ensuring that space remains accessible and safe for current and future missions.

Role Description

Cosmoserve Space is looking for a Computer Vision Engineer who is working on perception systems for spacecraft and space-object relative navigation. He/She will design and build algorithms that estimate the 6-DoF (position and orientation) of non-cooperative or known space objects for autonomous rendezvous, proximity operations, and docking. This role sits at the intersection of classical computer vision (detection, geometry, calibration) and modern deep learning, with heavy use of synthetic data generated in simulation environments.

Key Responsibilities

• Building vision-based relative navigation pipelines for object detection, instance/semantic segmentation, and pose estimation of rigid space objects.
• Working with PnP algrithms and RANSAC, epipolar geometry, and homography-based pose recovery techniques, along with feature descriptors such as SIFT, ORB, etc
• Applying data augmentation and domain randomization strategies to close the sim-to-real gap across illumination, albedo, distance, and attitude variations.
• Handling classical image processing tasks, including filtering, morphological operations, and edge/contour and blob detection, alongside learned features.
• Generating and curating photorealistic synthetic datasets in Blender, using procedural scenes, domain randomization, and automated ground-truth pose/segmentation export.
• Working across stereo and monocular camera geometry, including depth estimation, triangulation, and multi-view sensor fusion.
• Managing camera calibration workflows, including intrinsic/extrinsic parameters, lens distortion modeling, and stereo rectification, for single and multi-camera rigs.
• Optimizing and deploying models on embedded platforms such as NVIDIA Jetson, using TensorRT, ONNX, and CUDA for real-time inference, including profiling and quantization.
• Collaborating with GNC (Guidance, Navigation & Control) and simulation teams to feed pose and uncertainty estimates into the navigation filter.

Required Skills & Qualifications

• 2+ years of relevant industry or research experience with a master’s degree, or 4+ years of relevant industry or research experience, in Computer Vision, Robotics, Computer Science, or a related field.
• Strong foundation in computer vision and deep learning: CNN backbones, detection architectures, segmentation architectures, pose estimation networks.
• Solid grasp of projective geometry: PnP with RANSAC and Experience in applying data augmentation and domain randomization for robust, generalizable models.
• Strong command of classical image processing using OpenCV (filtering, morphology, feature extraction, edge/contour detection).
• Working knowledge of stereo and monocular camera geometry, depth estimation, and rectification.
• Proficiency in Python, PyTorch, OpenCV, and NumP y, with hands‑on experience in model training, evaluation, and profiling.

Nice to Have

• Hands‑on camera calibration experience (single or multi‑camera, intrinsic/extrinsic).
• Prior exposure to spacecraft navigation, robotics perception, or similar safety‑critical vision applications.
• Familiarity with sensor fusion and state estimation (e.g., Kalman filtering) in a navigation context.