Abhisek Kundu is an expert in machine learning and is currently working at Intel Labs India.

He received his doctorate in computer science from Rensselaer Polytechnic Institute.

INDIAai interviewed Abhisek Kundu to get his perspective on AI.

How did you get started with AI/ML?

I learned about AI and ML in my undergrad classes, where I was intrigued by the concept of intelligence's 'artificial' nature and the related computational techniques. Although, the real fascination for AI came through the realization of its potential to revolutionize human civilization. This situation made me choose AI/ML as a career.

What initial challenges did you face as a researcher, and how did you overcome them?

My initial days as a researcher were full of unguided enthusiasm and excitement. The first challenge was to find a proper direction of research to pursue. The other challenges were to develop an in-depth understanding of the top research papers, stay updated with the latest research results and contribute to the research community through innovation. 

Initially, I focused on building solid fundamental knowledge of the related subjects and improving the breadth of my understanding through coursework. Afterwards, I read more and more research papers to stay updated about the research trends and the solutions/breakthroughs to unsolved problems.

How does an AI/ML Scientist at Intel Labs spend a typical day?

A typical day for AI/ML scientists at Intel Labs involves coding and running experiments, discussing the latest research results and project planning/alignment so that we move towards achieving key technical goals and outcomes. Meanwhile, we take part in activities for team bonding and socialization. Overall, every day we strive to maintain Intel's culture and values to keep Intel a great place to work.

What, in your opinion, makes a sound machine learning engineer into a great one?

A sound ML engineer should possess proficiency in coding, fundamental technical knowledge, and the ability to deliver on time. Apart from such technical expertise, entrepreneurial and leadership skills to drive the results to make a societal impact make an ML engineer great.

What, in your opinion, will be the next major trend in AI research during the next decade?

The major trends of AI could be:

• Autonomous driving: trial runs are already happening in big cities. 
• AI-driven healthcare: breakthroughs like AlphaFold can rapidly expedite this field.
• Design of large/efficient/scalable models for multi-modal learning to move toward Artificial General Intelligence
• AI-driven Augmented Reality: Metaverse
• AI-driven chip design: the success of the above significant trends is contingent on the availability of massive computational resources. AI has the potential to revolutionize various stages of the chip design process.

Who is your role model in AI/ML?

The real success of AI/ML comes through its impact on human civilization, which is dependent on the progress of scalable and robust algorithms, software frameworks and tools, and efficient hardware platforms. The top contributors to such fields are our heroes. 

We have veteran stalwarts such as Geoffrey Hinton, Yann Lecun, Yoshua Bengio, and Jurgen Schmidhuber as role models. On the other hand, young researchers such as Ian Goodfellow (creator of Generative Adversarial Networks) and Kaiming He (creator of Resnet) inspire younger researchers to make a mark in the field. 

Creating software frameworks, such as PyTorch and Tensorflow, helped developers rapidly progress on the algorithmic front. Similarly, the availability of scalable hardware platforms to train/deploy such large-scale AI models is critical for progress in AI. Such software/hardware design happens via collaboration and team effort. Such teams are also our role models.

AI/ML stalwarts in India like Prof Sankar Kumar Pal, Prof Sanghamitra Bandyopadhyay, and Prof Balaraman Ravindran are our role models.

What, in your opinion, are the main areas where Indian researchers are falling short?

Historically, science and technology research in India always faced initial challenges from a lack of equipment/resources. But then we have great scientists and leaders who rose to the occasion and inspired generations of researchers to make rapid progress in those fields to place India amongst the best (take ISRO as an example). AI research in India also faced similar challenges. We lack extensive computational resources critical for massive-scale AI to impact society. However, these problems are now getting mitigated through global-scale collaboration between industry and academia. Leading AI companies are investing heavily in AI research in India. These insights highlight that we never had a shortage of talents. With such investments in computational and human resources, I am confident that in the following decades, India can make a significant mark on the global map of next-generation AI. 

What advice would you give someone who wants to work in AI research? What are the best ways to get ahead?

Here are some of the advice I followed:

1. Focus on building solid fundamental knowledge of ML concepts, optimization and statistics, and coding proficiency. 
2. Read the latest research papers on the problem of interest to stay updated.
3. Identify the open issues (both algorithmic and applied) and brainstorm to solve them.

The heroes are the ones who make our life easier through their contributions. To develop a scientific and entrepreneurial mindset to impact society using AI as a tool.

Could you give me a list of important research articles and books?

Some classic books I follow are:  

1. Pattern Recognition and Machine Learning by Christopher Bishop (for general ML concepts)
2. Artificial Intelligence: A Modern Approach by Stuart Russell and Peter Norvig (for general AI concepts)
3. Deep Learning by Ian Goodfellow and Yoshua Bengio, and Aaron Courville (for Deep Learning concepts)
4. Algorithms for Optimization by Mykel J. Kochenderfer and Tim A. Wheeler (for ML-specific optimization)

However, I spend much more time reading papers and articles published in top ML conferences and journals. The best papers of these conferences often provide breakthrough solutions and set the next big trend in AI/ML. Some top ML conferences are NeuRIPS, ICML, ICLR, AAAI, UAI, and CVPR. Also, survey papers on a specific research problem can be an excellent handbook for understanding the pros and cons of many existing approaches to such issues.