Computer Vision Engineer

SKEYEBOTS; Bengaluru; On-site; INR 1,800,000 - 2,600,000; Full time

Benefits offered by this job

Ownership of perception modules

Collaboration with autonomy, controls,

Well-equipped robotics lab

Mentorship and publication support

Health insurance and ESOP

Job summary

SKEYEBOTS in Bengaluru, India is hiring a Perception Engineer to turn perception research into reliable, real-time onboard systems for UAVs. You will work on camera, LiDAR, and radar sensing, develop CV pipelines, and build UI to monitor detections.

The role involves on-site work, closely collaborating with state estimation, flight control, planning, and hardware teams, with emphasis on low-latency inference and ROS integration.

Qualifications

• 2-3 years building computer vision or perception systems.
• Experience deploying on real hardware (UAV preferred).
• Strong Python and modern C++ (C++14/17) programming.
• Hands-on deep learning with PyTorch for detection/segmentation.

Responsibilities

• Design, train, and deploy models for aerial 2D/3D object detection from camera, LiDAR, and radar.
• Design and implement real-time CV algorithms for flight, including visual odometry and SLAM.
• Build radar perception pipelines for detection and tracking in low visibility.
• Develop multi-sensor fusion (camera-radar, camera-LiDAR, visual-inertial) for robust flight understanding.
• Own sensor calibration and time synchronization across cameras, radar, LiDAR, and IMU.

Skills

Python

C++

PyTorch

OpenCV

ROS 2

Git

Linux

Education

B.Tech/M.Tech/MS in CS/EE/Robotics

Tools

ROS 2

PX4/ArduPilot

MAVLink

NVIDIA Jetson

TensorRT

CUDA

OpenCV

PyTorch

Job description

SKEYEBOTS | Bengaluru, India | Full-Time | 2-3 years | Immediate joiners prioritised

About SKEYEBOTS

SKEYEBOTS builds AI-driven autonomous systems for unmanned platforms across aerial, ground and maritime environments, where perception failures are not an option.

Role Overview

We are hiring a Perception Engineer with 2-3 years of experience to build the sensing stack that lets our UAVs see and understand their surroundings, using cameras, LiDAR and radar. You will take perception models and pipelines from research prototypes to reliable, real-time systems running on board drones in flight. You will also build the visual interfaces operators use to monitor them.

This is an immediate opening. Candidates who can join right away will be prioritised.

At a Glance

Experience: 2-3 years (industry or post-Master's research)

Location: Bengaluru, India (on-site; lab and flight testing, with travel as required)

Employment: Full-time

Reports to: Perception Lead

Team: UAV autonomy. You'll work closely with state estimation, flight control, planning, controls, UI, and hardware.

Key Responsibilities

• Design, train, and deploy models for aerial 2D/3D object detection, segmentation, and tracking from camera, LiDAR, and radar data.
• Design and implement classical computer vision algorithms for real-time flight: visual odometry, SLAM, feature matching and optical flow.
• Build radar perception: process mmWave/FMCW radar data (range-Doppler maps, point clouds, CFAR detection) for detection and tracking in low visibility.
• Build multi-sensor fusion pipelines (camera-radar, camera-LiDAR, visual-inertial) for robust scene understanding in flight.
• Develop perception for UAV autonomy: obstacle detection and avoidance, landing-zone detection, terrain mapping, and detect-and-avoid of other aircraft.
• Own sensor calibration and time synchronisation across camera, radar, LiDAR and IMU, including vibration and motion effects on board. Work with the hardware and controls teams on camera and sensor selection.
• Optimise models and CV pipelines for real-time inference on size-, weight- and power-limited edge hardware (e.g., NVIDIA Jetson) using TensorRT, quantisation and profiling for latency, memory and power.
• Integrate perception into the ROS 2 and PX4/ArduPilot stack with clean interfaces to estimation and planning. This covers sensor drivers, message pipelines, custom nodes, and vision-aided navigation and state estimation on PX4.
• Write performance-critical code in C++, and prototype and validate algorithms in Python.
• Contribute to QML/Qt-based UI/HMI components for visualising detections, telemetry overlays, and operator dashboards.
• Build data workflows: flight data collection, labelling guidelines, dataset curation and versioning.
• Define metrics and run evaluations (mAP, IoU, MOTA, latency) in simulation (Gazebo/SITL), on flight logs, and in field trials. Debug failures on real flights.
• Write unit and integration tests and maintain CI pipelines for perception modules.

Required Qualifications

• B.Tech/M.Tech/MS in Computer Science, Electronics, Aerospace, Robotics, or a related field.
• 2-3 years building computer vision or perception systems, with at least one system deployed on real hardware (UAV preferred).
• Strong Python and modern C++ (C++14/17), including writing performance-critical code.
• Hands-on deep learning with PyTorch: training, fine-tuning, and evaluating detection or segmentation models.
• Working knowledge of radar signal processing: FMCW principles, range-Doppler-angle estimation, CFAR and clustering.
• Solid grounding in multi-view geometry and CV fundamentals: camera models, calibration, projections, epipolar geometry, feature detection, optical flow, and 3D coordinate transforms. Hands-on with OpenCV.
• Working experience with ROS or ROS 2 (nodes, topics, services, tf/tf2, launch files).
• Experience with multi-sensor fusion and tracking (Kalman filters, data association).
• Comfortable with Linux, Git, and debugging on embedded or flight hardware.

Preferred Qualifications

• Hands-on with PX4 or ArduPilot, MAVLink, offboard control, and flight testing (awareness of DGCA drone rules is a plus).
• Experience with automotive/aerial mmWave radars (e.g., TI AWR/IWR series) and radar-camera fusion models.
• Deep-learning-based radar perception on raw ADC data, radar cubes, or radar point clouds.
• Deployment with TensorRT, ONNX Runtime, or CUDA on Jetson-class devices, or other NPUs.
• Exposure to visual-inertial odometry or SLAM (VINS-Fusion, ORB-SLAM, RTAB-Map, factor graphs) for GPS-denied flight.
• Experience building UI components with QML/Qt.
• Multi-modal sensor fusion including thermal/IR alongside RGB, stereo/depth and LiDAR, plus IMU and GPS fusion.
• Familiarity with simulators (AirSim, Isaac Sim, Gazebo, jMAVSim), SITL/HITL, and synthetic aerial data generation.
• Experience with TensorFlow in addition to PyTorch.
• Experience setting up CI pipelines and test frameworks for perception or robotics software.
• Publications at robotics, vision or radar venues (ICRA, IROS, CVPR, RA-L, IEEE RadarConf) or strong open-source contributions.

Tools and Tech Stack

Deep learning: PyTorch, TensorFlow (plus), ONNX, TensorRT

Vision and 3D: OpenCV, Open3D, PCL

Radar: mmWave/FMCW radar SDKs, range-Doppler processing, CFAR, DBSCAN

UAV and robotics: ROS 2, PX4 / ArduPilot, MAVLink, AirSim / Gazebo / jMAVSim, SITL/HITL

UI: Qt / QML

Sensors: RGB, thermal and stereo cameras, mmWave radar, 3D LiDAR, IMU, GNSS

Hardware: NVIDIA Jetson Orin, flight controllers (Pixhawk), GPU workstations

Infra: Linux, Docker, Git, CI, experiment tracking (W&B or MLflow)

What We Offer

• Ownership of perception modules running on real robots, from design to field deployment. Your work goes into real, operational autonomous systems, not just simulation.
• Direct collaboration with the autonomy, controls, and hardware teams on deployed platforms.
• Access to a well-equipped robotics lab, sensor rigs, and GPU compute.
• Mentorship from senior researchers and engineers, with support to publish and attend conferences.
• Competitive compensation + ESOP allocation, health insurance and a learning budget.
• A fast track to greater ownership as the perception team scales.

SKEYEBOTS is an equal opportunity employer. We evaluate candidates solely on capability, experience, and drive.