Engineer III - Computer Vision & Machine Learning

What This Role Offers

- Prospect to develop learning-based vision-aided navigation systems for autonomous UAVs.
- Hands-on exposure to real-world flight testing, dataset collection, and field validation of AI-driven autonomy.
- Ownership of algorithms from dataset preparation and training through geometric validation and embedded deployment.
- Close collaboration with perception, navigation, estimation, embedded and flight-control teams.
- An aggregated scope for R&D; and software development.

About The Role The Mid-Senior Computer Vision & Machine Learning Engineer is a hands-on role focused on building, adapting, validating and deploying computer-vision and machine-learning pipelines for visual navigation. The role requires strong practical experience in computer vision, deep learning, and real-time deployment on embedded platforms.

This is not a generic image classification or object detection role. The engineer must understand how visual algorithms produce measurements that can be consumed safely by a navigation or state-estimation system.

The engineer is not expected to independently architect the complete navigation stack but must be capable of owning individual CV/ML pipelines and delivering estimator-ready outputs.

This role is ideal for engineers who enjoy taking algorithms from research to real-life deployment.

Key Responsibilities

- Develop and improve learned local-feature detection, description and matching pipelines for vision-aided navigation.
- Generate positive and negative correspondence pairs using known poses, geometry, depth, synthetic transformations or pseudo-labels.
- Improve feature repeatability, spatial distribution, descriptor discriminability and matchability across changing viewpoints and altitudes.
- Apply geometric verification and robust outlier rejection before using learned correspondences for pose estimation.
- Build and manage datasets and data pipelines, including data collection during flight tests, annotation, augmentation, training, and evaluation workflows.




- Ensure robust runtime behaviour under real-world conditions such as motion blur, lighting variation, and sensor noise.
- Export and deploy models using ONNX, TensorRT or equivalent deployment runtimes.
- Distinguish model confidence from statistical measurement covariance.
- Integrate vision and ML outputs with robotics and autonomy stacks to enable downstream tasks such as navigation, obstacle avoidance, mapping, and decision-making.
- Document system designs, experiments, model performance, and deployment workflows to support maintainability and knowledge sharing.

Minimum Qualifications

- Bachelor’s or Master’s degree in Robotics, Electrical/Electronics Engineering, Computer Science, Aerospace Engineering, Mathematics or a related field.
- 2+ years (Master’s) or 4+ years (Bachelor’s) of industry or research experience in computer vision, machine learning, or robotics-related roles.; demonstrated technical depth and ownership are more important than a strict year count.
- Strong Python and working C++ proficiency.
- Practical experience with PyTorch or another modern deep-learning framework.
- Practical experience with OpenCV and classical multi-view geometry.
- Experience training or fine-tuning computer-vision models.
- Solid understanding of Image Processing (denoising, deblurring, contrast enhancement, and low-light image improvement)
- Experience with keypoint detection, descriptor extraction, feature matching and geometric verification.
- Understanding of pinhole, fisheye, and other relevant camera models; camera calibration; projection; homography; epipolar geometry; PnP; optical flow; and RANSAC.
- Understanding of coordinate frames, rotations, quaternions, timestamps and uncertainty.
- Understanding of camera and lens properties - shutter, aperture, etc.




- Working knowledge of Linux, Git and software debugging practices.
- Experience with ROS1 or ROS2.

Mathematical Expectations Candidates Must Demonstrate An Applied Understanding Of

- Matrix operations, coordinate transformations and linear systems.
- Linear Algebra - Eigenvalues, eigenvectors, SVD, rotation matrices, quaternions and Jacobians.
- Probability - Conditional probability, Bayes’ rule, probability distribution.
- Statistics - Random process, covariance, correlation, and first-order uncertainty propagation.
- Confidence intervals, hypothesis testing and statistical evaluation.
- Numerical conditioning and optimization (linear and convex).

Mathematical understanding should be demonstrated through vision and navigation problems, not only theoretical definitions.

Preferred Qualifications

- Experience with VIO, visual odometry, visual localization or SLAM.
- Experience with aerial, satellite, EO or thermal/IR imagery.
- Applied knowledge of distillation using model outputs, descriptors, intermediate features, correspondences or downstream pose supervision.
- Experience in pruning or architecture simplification.
- Experience in diagnosing unsupported operators, memory-transfer bottlenecks and CPU–GPU synchronization issues.
- Experience with Jetson Orin NX, TensorRT, CUDA or NVIDIA profiling tools.
- Experience with ArduPilot, PX4, MAVLink or autonomous UAV systems.
- Experience with Docker and reproducible ML environments.
- Familiarity with estimator interfaces, innovation gating or Kalman-filter concepts.

Working Hours

- Standard working hours are 9:30 AM to 6:30 PM, Monday to Friday.
- Field-testing activities may require early-morning or extended hours, depending on mission requirements.

Compensation Range

- Competitive compensation aligned with industry standards, including performance-based incentives.
- Exact salary ranges can be customised according to experience.

Advantages

- Comprehensive health insurance.
- Professional development support.
- Detachment allowance.

📌 Engineer III -CVML (Bengaluru)
🏢 NewSpace Research & Technologies
📍 Bengaluru