An Invaluable Learning Experience with Real-world Opportunities

The machine learning (ML) bug bit Pang Hui En when she joined Grab for a six-month stint as a data science intern. “I was using ML for activity recognition and GPS (global positioning system) correction and I was intrigued by the possibility of other applications,” she said. 

But with a degree in Environmental Studies and her previous job as GIS research assistant at the National University of Singapore (NUS), she did not have the confidence to embark on a career in data science.

“I felt I lacked the necessary skillsets as I was not from a technical background like computer science or statistics,” she said. “I also did not have experience with model deployment, nor did I know how to build an end-to-end ML system as I was mostly doing model training in my Grab Internship.”

That would soon change. A senior at the National University of Singapore, a chemical engineer who had made a job switch to become a data scientist, introduced her to the AI Apprenticeship Programme (AIAP)®. “He encouraged me to try for this programme if I was serious about becoming a data scientist,” she said.

AIAP presented her with the opportunity to pick up software engineering and model deployment skills and to build an end-to-end machine learning system.

A diverse cohort

To prepare for the programme, she dug into the AI Apprenticeship Field Guide and worked on Kaggle challenges published by the online data science community.

Still, there was a niggling concern that she lacked the necessary technical competency or exposure to data science.

But she needn’t have worried.

On starting the AIAP, she found that most of the apprentices did not have a computer science degree either. She got to interact with biomedical engineers, economics and business majors, and a whole lot of other people from diverse backgrounds, and this only enriched the learning experience. “Having a diverse background is an advantage because you bring your own domain knowledge to the table and formulate a problem differently.”

The apprentices were also asked about their areas of interest and matched to projects that would enable them to pick up the skills that they wanted to acquire.

Safe environment to learn

Hui En, who is currently in AIAP Batch #6,  joined the AISG team working on Finepose, a project that uses computer vision models in real-world applications, for example, pose estimation models for social-distancing. 

When team members found out that she was keen to pick up software engineering skills, they gave her the opportunity to learn.

One of her first tasks was to refactor a portion of the codebase that previous batches of apprentices had worked on. In order to do this, she had to quickly understand the code, its structure and how it works.

“During my first refactoring, I made several mistakes but thankfully, AIAP is a safe environment to learn,” she said.  

Her mentor gave her useful tips and advice on writing cleaner and more efficient code, and this has helped her to become a more all-rounded data scientist.

An unbeatable experience

Hui En is in charge of model training and the training pipeline, and a challenge for the team is to build accurate and lightweight models for deployment. This is something that she has been looking forward to, and the experience has been an eye-opener for her.

“I used to think that model accuracy was the most important metric or indicator of how good the model was. However, working on Finepose, I realised that there is a ton of resource constraints when deploying a model,” she said. 

“There is always a trade-off between model accuracy and size/inference time. This is important for real-world applications and deployment, and I learnt how to build models that try to achieve a good balance between both.”

No doubt the nine-month AIAP is a rigorous programme, but having the right attitude makes the learning process easier, said Hui En. And the opportunity to apply ML skills on real-world applications is a really valuable experience that online courses cannot provide. Summing up the experience, she said, “It has been an extremely fulfilling journey thus far.”