Landing a job opportunity is one challenge. Being truly ready for it is another.
Haidi Azaman learnt this lesson during his first data science internship. As a mechanical engineering undergraduate who wanted to make a switch, he realised how steep the climb would be.
“I wasn’t at a level where I could do data science yet,” he recalls. “I didn’t even know basic Python packages like pandas, which is a software library used to organise and analyse data.”
The role was with a maritime start-up developing algorithms to predict ship arrival times. It was one of only two companies that responded to his 200 cold LinkedIn messages.
Faced with a gap in his skills but refusing to give up after finally getting a foot in the door, Haidi spent several months developing his coding skills, learning how to clean datasets, uncover patterns and use data to make predictions – the core building blocks of data science.
His interest in data science had taken root midway through university, when he realised he was more drawn to software than hardware. A casual conversation with a senior about machine learning and how neural networks mimic the human brain sparked his interest.
“What fascinated me was how powerful software could be,” he says. “You may be one person behind a computer, but what you build can create real impact.”
His engineering background, he realised, had quietly prepared him for the transition. The training had taught him how to break complex problems into parts, test assumptions and iterate – habits that would later anchor his work in data science. More importantly, it gave him the confidence to act once his interests became clear.
In his third year, Haidi received the A*STAR Research Internship Award (ARIA), which gives undergraduates the opportunity to sharpen their skills and discover their passion in scientific research.
During his one-year internship, he worked on cognitive AI research, from conducting experiments to analysing results. After graduating from the National University of Singapore (NUS) with a bachelor of engineering (mechanical engineering) in 2023, Haidi applied for the SG Digital Scholarship (Postgraduate) – offered by the Infocomm Media Development Authority – to pursue a master of science in data science and machine learning at NUS.
What stood out to him was the flexibility of the bond. Scholars can serve in approved tech or media roles across Singapore, rather than being tied to a single organisation.
“That was my ‘aha’ moment,” he says. “It gave me the freedom to define my own career path in a tech company.”
“What fascinated me was how powerful software could be. You may be one person behind a computer, but what you build can create real impact.”
– Haidi Azaman, recipient of the SG Digital Scholarship (Postgraduate)
Haidi also interned at the Singapore office of ride-hailing firm Gojek, where he worked on strengthening facial-verification systems used by drivers.
Some of them, he learnt, would attempt to trick the system by holding up printed photos to the camera.
The solution did not come from data, but from a simple human insight grounded in observation.
“A printed photo doesn’t blink,” Haidi explains.
Together with his team, he successfully trained an image-classification model to detect a subtle, involuntary human behaviour: an eye blink.
It became an additional signal the system could use to verify that a real person – not an image – was in front of the camera.
That experience showed him that even as AI becomes more autonomous, progress still begins with intentional human input.
As part of IMDA’s scholars programme, he also had the opportunity to visit the offices of TikTok and Amazon Web Services. “There is a lot of support for the additional training courses that you want to take,” he says.
In 2024, Haidi joined SAP Labs Singapore in a full-time role as an AI scientist. He works on advanced data systems that help organisations retrieve information accurately and reliably.
For all the sophistication of today’s machine learning models, Haidi is quick to stress where responsibility still lies.
“Before any model is trained, someone has to understand the data,” he says. “If humans don’t do that carefully, the system will give confident but wrong answers.”
In the age of AI, he believes human judgement matters more than ever.
