Computer Vision Engineer — Terminal Guidance (EO/IR)

About the Role:

Green Aero Propulsion is building India's first indigenous transonic unmanned strike platform, AgniPankh. We are hiring a Computer Vision Engineer to own the EO/IR-based terminal guidance pipeline — from sensor fusion and target acquisition to real-time tracking and precision strike handoff. You will design, train, and deploy deep-learning models that recognise and lock onto targets in GPS-denied, high-speed, low-altitude flight regimes where latency and robustness are non-negotiable.

Location: KIADB Aerospace Park, Devanahalli, Bengaluru (on-site)

Employment type: Full-time

Reports to: Head of Avionics

Key Responsibilities:

Develop and optimise real-time object detection, classification, and tracking algorithms for EO/IR sensor payloads operating at transonic speeds

Design the terminal-guidance computer-vision pipeline: target acquisition, lock-on, mid-course update, and handoff to the flight controller for precision terminal manoeuvre

Build, curate, and augment training datasets from UAV-captured aerial imagery, satellite imagery, thermal imagery, and synthetic data (domain randomisation, sensor-noise injection, clutter generation, weather/lighting variation)

Train and compress deep-learning models (YOLO variants, transformer-based detectors, Siamese trackers) for edge deployment on embedded GPU/FPGA hardware under strict size, weight, and power (SWaP) constraints

Integrate vision algorithms with the TERCOM-based GPS-denied navigation stack and the 6-DOF digital-twin simulation environment for hardware-in-the-loop (HIL) testing

Implement sensor-fusion techniques combining EO, IR, and INS data to maintain target track through clutter, countermeasures, and environmental obscurants

Define and run model-validation campaigns: precision/recall benchmarks, latency profiling, Monte Carlo miss-distance and CEP analysis, and flight-test correlation

Collaborate with the avionics, aerodynamics, and propulsion teams to ensure vision-system requirements align with platform-level performance budgets

Stay current with advances in vision-based guidance, scene matching, and adversarial robustness; evaluate applicability to defence-grade systems

Must-Have Qualifications:

B.Tech / M.Tech / Ph.D. in Computer Science, Electrical Engineering, Aerospace Engineering, or a related discipline with a strong computer-vision focus

1+ years of hands-on experience in object detection, recognition, and tracking using deep-learning frameworks (PyTorch, TensorFlow, or equivalent)

Demonstrated work on aerial/UAV imagery — familiarity with oblique viewing angles, scale variation, motion blur, and low-contrast IR scenes

Strong programming skills in Python and C/C++; comfort with Linux, Git, Docker, and CI/CD pipelines

Solid understanding of image-processing fundamentals: camera models, lens distortion, homographies, multi-spectral image registration

Familiarity with sensor-fusion concepts (EO + IR + INS) and Kalman/extended-Kalman or particle-filter-based tracking

Working knowledge of ROS/ROS 2 for perception-pipeline integration in robotic or UAV systems

Indian citizen with ability to obtain necessary security clearances for defence projects

Preferred Qualifications:

Prior experience in defence, missile-guidance, or loitering-munition vision systems

Hands-on work with scene-matching algorithms (TERCOM, DSMAC, or similar correlation-based guidance techniques)

Experience with synthetic-aperture or millimetre-wave radar image processing

Published research or patents in aerial object detection, visual SLAM, or adversarial robustness for safety-critical vision systems

Experience training and deploying models on embedded edge platforms (NVIDIA Jetson, Intel Movidius, Xilinx/AMD FPGA, or comparable)

Proficiency in model optimisation: quantisation (INT8/FP16), pruning, knowledge distillation, and TensorRT / ONNX Runtime deployment

Familiarity with MIL-STD or DO-178C/DO-254 certification processes for airborne software/hardware

Experience with GAN-based or diffusion-model-based synthetic data generation for rare-target augmentation

Proficiency with simulation environments such as AirSim, Gazebo, or FlightGear for vision-in-the-loop testing

What we offer:

A founding-team-level role in a DPIIT-recognised defence aerospace startup building category-defining indigenous platforms

Direct ownership of the terminal-guidance vision stack from research through flight test

Competitive salary with ESOP participation

Access to in-house engine test cells, avionics labs, and a 6-DOF digital-twin simulation facility at the HiTech Defence & Aerospace Park, Bengaluru

Opportunity to work at the intersection of AI and national security on systems that will see operational deployment

A fast-moving, engineering-first culture where your code flies

Seniority level: Entry level

Employment type: Full-time

Industries: Aviation and Aerospace Component Manufacturing