Getty Images

How an ML engineer cut his teeth in AI

Machine learning engineer Saurabh Agarwal talks up his career journey in artificial intelligence and what it takes for one to succeed in the field

This article can also be found in the Premium Editorial Download: CW Asia-Pacific: CW APAC – Tech Career Guide: Artificial intelligence

At 8am, Saurabh Agarwal can be found flipping through the pages of some latest artificial intelligence (AI) research paper over coffee to keep pace with the dynamic field.

Agarwal, a machine learning (ML) engineer at MavQ, an India-based AI platform company, has seen AI developments unfolding since the day he dipped his toes into the technology.

A computer science graduate from Jaipur in North India, Agarwal started fiddling with the power of data during a full-time internship in data engineering and cloud. He soon found himself cutting his teeth deeper into data analytics and building data pipelines for ML models.

Under the guidance of mentors such as Gaurav Kheterpal, a Salesforce trailblazer, Mulesoft specialist and multicloud expert, and with the support of AI communities, Agarwal cemented his grip in ML modelling.

For over two years, he learnt how to develop ML models from scratch and, later, got the hang of deploying them. Besides scaling models, he also handles models for converting paper documents into digital formats while working on deep learning. Alongside his work, he is finishing an executive course in AI and ML from the Birla Institute of Technology and Science, Pilani.

“The sheer joy of seeing what data can do is exciting. In the previous IT boom, we had a faint idea of what’s possible, but back then data was limited. Today, the exponential jump in data consumption has made models easy and resource-friendly,” says Agarwal.

“In the next five years, I see a lot of scope for MLOps [machine learning operations] and model productisation. It is very rewarding to be involved in scaling models at my company, and also making sure that they are not bulky or expensive. Keeping a good tab on costs and competitive edge makes this job thrilling every day,” he adds.

Machine learning can be a lucrative career in India. According to Agarwal, an average ML engineer’s salary package can range from Rs. 15-20 lakhs (US$18,000-24,000) per annum, with higher level domains commanding Rs. 60-70 lakhs per annum as one gets more proficient.

Besides MLOps, which involves productisation and optimisation of ML models, applied ML and advanced modelling are other hot domains. But those who are clear and sharp about their specific domain will do the most, according to Agarwal.

“Anyone with a good command of software engineering will find it easy to get in. One has to be comfortable and cognisant about data – that’s the basic trait. Before you get into ML modelling, an appreciation and a knack for how data affects business is paramount in this profession,” he says.

“AI is very vast. Understand your own domain. Do not hype it or dismiss it. It will not take away all jobs. The calculator never did. We need to adapt to AI and make it augment us. It’s that simple”
Saurabh Agarwal, MavQ

He also points out the importance of explainability and the ethics of modelling: “One has to be able to ask, ‘Why is this model telling us this?’ and be aware of its implications.”

Agarwal spends most of his day leading a team of 15 to 20 people to productise ML models. He is also involved in benchmarking of models against the top ones in the industry. “We have to constantly check how accurate, how good and how fast we are vis-a-vis the top crust in the industry. Right now, we are either on par or better than the big names,” he says.

Agarwal is confident that AI will only amplify data’s potential for business in the future. “As models and deep learning get better, we will find AI working to our advantage. For that, scaling and executing models while confronting the ‘black box’ problem would be crucial,” he adds, referring to the challenge of understanding how decisions are made by some complex AI models.

Agarwal usually wraps up his day with a jog or a badminton match. On the track, he knows when to sprint and pause, something that comes naturally to a good AI professional.

His advice to ML and AI aspirants is twofold: choose your domain well and have an innate love for data. “AI is very vast. Understand your own domain. Do not hype it or dismiss it. It will not take away all jobs. The calculator never did. We need to adapt to AI and make it augment us. It’s that simple,” he says.

Read more about AI in APAC

Read more on Artificial intelligence, automation and robotics

Data Center
Data Management