Adelaide, South Australia

Mohsi Jawaid

Computer vision researcher working on vision-based orbital guidance, navigation and control — teaching spacecraft to work out where they are from what their cameras see, even when the lighting is brutal.

University of Adelaide Australian Institute of Machine Learning
Get in touch GitHub LinkedIn
Mohsi Jawaid

About

I'm a computer vision researcher at the University of Adelaide, working with the Australian Institute of Machine Learning on perception for spacecraft. My PhD tackled the sim2real domain gap: models for satellite pose estimation have to be trained on synthetic imagery, because real in-orbit data with ground truth barely exists — and they tend to fall apart the moment they meet a real camera.

Most of my work leans on event cameras (neuromorphic sensors that report per-pixel brightness changes instead of frames). Their enormous dynamic range makes them far more robust to the glare, over-exposure and lens flare that wreck conventional RGB in orbit. Lately I've been fusing the two sensor types, and putting pose estimation onto neuromorphic processors.

Before research I was a software engineer at Maptek, building machine learning features for mining software, and I've taught object-oriented programming at Adelaide.

240+ Citations
5 h-index
8 Publications

Experience & education

  • Grant-Funded Researcher (B) Current
    School of Computer Science and IT, University of Adelaide

    Vision-based orbital guidance, navigation and control. Event-based and event–RGB spacecraft pose estimation, hardware-in-the-loop experiments, and neuromorphic deployment.

  • PhD, Computer Science Thesis submitted 2026
    University of Adelaide · Australian Institute of Machine Learning

    Bridging the synthetic-to-real domain gap for spacecraft pose estimation using event sensing — including the SEENIC dataset and test-time self-supervision methods.

  • Software Engineer Previously
    Maptek

    Joined through an internship and stayed on, developing machine learning features for mining and geospatial software.

  • Teaching — Object-Oriented Programming —
    University of Adelaide

    Tutoring and teaching support for undergraduate programming courses.

  • Bachelor of Computer Science (Advanced) 2016 – 2018
    University of Adelaide

    Graduated with multiple prizes and Executive Dean's Awards for academic excellence.

What I work on

Event-based space vision

Neuromorphic sensors for spacecraft pose estimation — high dynamic range perception that survives glare, blooming and eclipse transitions.

Sim2real domain adaptation

Closing the gap between rendered training data and real sensor data, including test-time self-supervision with certifiability guarantees.

Sensor fusion

Combining event and RGB streams so each covers the other's blind spots: resolution and texture from RGB, dynamic range and latency from events.

Synthetic data & HIL testing

Virtual production pipelines and robotic hardware-in-the-loop rigs for generating realistic, ground-truthed orbital navigation datasets.

  • Computer vision
  • Deep learning
  • Event cameras
  • Pose estimation
  • Domain adaptation
  • Generative models
  • Python
  • PyTorch
  • C++

Selected publications

Full list on Google Scholar.

Code & projects

space-event-rgb-fusion

Multimodal event–RGB fusion for spacecraft pose estimation under harsh orbital lighting.

Python

event-pose-certification

Test-time certifiable self-supervision for event-based satellite pose estimation — the reference implementation for the IROS 2024 paper.

Python

Awards & honours

  • Top rankings, ESA Kelvins Satellite Pose Estimation Challenge — as part of Team TangoUnchained
  • Australian Computer Society Prize (2018)
  • Maptek Prize in Computer Science
  • Executive Dean's Awards — multiple, for academic excellence
  • Governor's International Award
  • Oracle User Group Prize