Let's talk robotics with Ragja Palakkadavath

Joining me today is Ragja Palakkadavath. Ragja is a second-year PhD student at Deakin University’s Applied Artificial Intelligence Institute (A2I2) where she is working on creating AI systems that can perform out-of-distribution generalization.

Undertaking fundamental research in AI, Ragja is investigating how to create more efficient and practical artificial intelligence systems. As she explains, AI systems don’t do well when they encounter data they weren’t trained to recognise; for example an autonomous car crashing because it failed to recognize a kangaroo on an Australian road, having been trained to recognize deer, or being able to recognise and read road signs in the USA but not in New Zealand. In another example, AI is increasingly being used to scan and detect disease in medical imaging, however the huge variation in data provided by patients and even how difference hospitals carry out imaging makes these systems less effective. Imagine if there was a single model that could work for all patients, and all doctors in all hospitals. Developing more robust AI models holds to promise of solving these all too real challenges.

 Ragja says none of this feels like work and she is never bored. When she can turn her mind away from the world of AI, she enjoys reading, cooking and travelling.

 It has been quite a journey so far. At age 10, Ragja left behind village life in Kerala, India, when her family moved to Muscat in Oman. The international multicultural community of the Middle East nation was a big change but one that ultimately gave her the confidence and independence to move to Australia in 2022.

 Discovering a love of coding at high school, Ragja had made up her mind at age 12 that she would have a career as a computer science engineer. Returning to Kerala, Ragja studied at the National Institute of Technology Calicut, and began working for one of the leading travel technology companies, Amadeus. Using machine learning to track passenger movements through airports she became fascinated with why some models were more effective than others and undertook her Masters, producing the thesisBayesian Generative Adversarial Nets with Dropout Inference in 2020.

 She says the beautiful beaches, scenery and welcoming people of Australia remind her of Oman.

Reach out to Ragja at s222101652@deakin.edu.au or connect with her on LinkedIn!