Looking for a job

Pretha Sur

@prethasur

Message

I'm building lightweight onboard AI for real-time SAR satellite perception and geospatial monitoring.

Australia

Message

What I'm looking for

I'm looking to build scalable geospatial AI systems for meaningful Earth-feature monitoring, particularly across maritime, infrastructure, sustainability, and climate-resilience applications.

I'm a PhD researcher at Deakin University and SmartSat CRC building lightweight AI for onboard SAR and Earth-observation analytics. My work targets real-time ship detection, maritime awareness, and scalable monitoring under satellite compute and downlink constraints.

I develop compact YOLO- and RT-DETR-style detectors, hardware-aware neural architecture search, and deployment-faithful evaluation using PyTorch, ONNX, TensorRT, CUDA-aware workflows, and NVIDIA Jetson Orin Nano. I measure accuracy alongside latency, tail behaviour, memory, quantisation sensitivity, and robustness in challenging SAR scenes.

As a visiting researcher at ESA Φ-lab/ESRIN, I'm developing operator-conditioned knowledge distillation for minimally processed SAR data. I've also delivered edge object-detection systems for factory automation and built 2D/3D detection and tracking pipelines for autonomous-driving simulation at IIT Bombay.

Experience

Work history, roles, and key accomplishments

Current

Visiting Researcher

Current

ESA Φ-lab

Jan 2026 - Present (8 months)

Developing an operator-conditioned knowledge distillation framework for SAR perception, where teacher and student models see different but physically related views of the same signal. Incorporating known SAR processing operators into the distillation objective.

Known Operator KD L0 SAR Perception Knowledge Distillation SAR Processing

Current

PhD Researcher

Current

Deakin University

Jan 2024 - Present (2 years 8 months)

Developing lightweight deep-learning pipelines for satellite-based SAR perception, focusing on ship detection and maritime domain awareness under onboard compute constraints. Leading hardware-aware NAS studies and deployment-faithful evaluation protocols.

Maritime Domain Awareness PyTorch Ultralytics Onnx TensorRT CUDA Knowledge Distillation Pruning

Computer Vision Intern

Indian Institute of Technology Bombay

Nov 2023 - Apr 2024 (5 months)

Implemented state-of-the-art perception pipelines for autonomous driving on CARLA/AirSim, including 2D/3D detection and tracking. Reproduced baselines and tuned data/augmentations for robustness.

Autonomous Driving 3D Detection Trackings CARLA Latency Trade Offs

ML Engineer

Resolute AI Software Pvt. Ltd.

Jul 2022 - Oct 2023 (1 year 3 months)

Delivered real-time object detection for factory automation and safety using YOLO, SSD, and MobileNet. Optimised pre/post-processing to achieve low-latency inference on edge devices.

YOLO SSD Object Detection Data Curation

Education

Degrees, certifications, and relevant coursework

Deakin University

Doctor of Philosophy, Artificial Intelligence / Remote Sensing

Activities and societies: Research affiliation with SmartSat CRC-supported space AI, satellite autonomy, and onboard Earth-observation analytics; supervised by Prof. Abbas Khosravi.

PhD candidate in Artificial Intelligence and Remote Sensing at IISRI, focusing on deployment-faithful lightweight AI for onboard SAR/EO analytics.