Sabin Hashmi is currently working on building a machine learning based trigger system for LHCb, CERN.
He works on machine learning and deep learning applications in high energy particle physics.
INDIAai interviewed Sabin to get his perspective on AI.
It’s admirable to see how a physics student becomes an AI researcher. What inspired you to pursue a career in AI?
My passion for science and technology is my main motivation to become an AI researcher. It’s always interesting to work in a field that is developing quickly. The AI community includes people from different fields, all curious and passionate about advances in technology and research. Physics has helped me with critical thinking, and AI-based research has helped me provide flexibility in research. As I progressed in the field of AI, it became more and more exciting and I decided to choose AI as my primary research area.
Rather than turning away from physics, my current research uses AI as the main supporting tool to solve a problem that was originally solved using traditional methods and transform it into a more efficient solution using computational physics and AI.
Tell us about your PhD in particle physics and machine learning. What are the main problem areas?
In the Large Hadron Collider at CERN, two beams of protons are accelerated to the speed of light (99.99%) and collide at the experimental sites intended to study subatomic physics. When particles collide at very high speeds, there is a shower of many subatomic particles inside. My PhD research mainly focuses on the particle reconstruction of rare decays resulting from the collision of protons at the experiment. It is developing advances in developing a triggering system using machine learning that selects the particle trails of rare decays in real time.
The main concern is the enormous amount of data generated in each collision and the development of a decision-making system that works in real time. Additionally, the physics behind the rare decays make it difficult to find the decays of interest, and ML helps identify and characterize traces of rare decays from other signals.
What are your current responsibilities and activities at CERN’s LHCb?
The Large Hadron Collider Beauty (LHCb) experiment mainly investigates CP violations and rare particle decays. My research area is mainly focused on developing a machine learning based pipeline for a software based trigger system. As a developer, responsibilities include keeping up to date with changes in the detectors and software code bases, presenting advances in research work in collaboration with experts, and taking on data collection shifts at the experiment site.
What are the advantages and disadvantages of doing research as an Indian scientist in Poland?
It’s great to work in a dynamic environment with a peer group of researchers who are curious about developments in research. Research abroad can open up more opportunities for you, including working closely with pioneers and subject matter experts. This approach would result in a large global footprint that could prompt the researcher to choose how to proceed with their research. Additionally, the research space offered by global research institutions will inspire you to advance as a researcher.
There are fewer disadvantages apart from leaving the homeland and people. But it is very promising to see that research and development in India in recent years has led to more opportunities within the country.
What are the top three societal problems you want to address with machine learning? Or do you want to make a unique appeal for one of them?
Machine learning is still in its infancy. I understand that the developments in AI are significant, but if we consider the potential of AI for a good cause, we’re not there yet. There are a variety of problems that we can solve with AI. Here the multi-domain expertise makes the AI more prominent. For example, AI-based advances in healthcare are novel cases that pique our interest. We’ve had healthcare developments in the past, but AI offered a different approach to solving the problems we’ve had for decades. Some of the latest developments in AI-based drug delivery, target identification for cancer cures, etc. top the list. Additionally, AI promises to predict weather and natural disasters more accurately than the traditional methods we’ve had in the past.
In short, the potential of AI is enormous. Rather than looking at it as a completely different area, I’d like to think of it as a supporting tool that helps researchers look at the problem from a different angle. We are progressively transforming the existing technology into a novel system embedded in ML and AI.
What do you think of India’s AI education system? How about the global situation?
The developments in India’s AI are on a global scale. There are long-term and short-term courses that you can enroll in in reputable colleges and other institutions in India. Depending on your career path, with a little research it is easy to decide on the course of study. In addition, there are many more ways to study courses from experts online.
In your opinion, what should be improved at Indian universities to advance AI? How should they proceed?
AI research is a vast field, so the course can be as broad as the field would be. Some courses revolve around the same concepts and how to develop them on the programming side of research. But AI is built on a solid foundation of math and statistics. Universities focused on advanced AI should be a place for students interested in learning and developing their problems and finding solutions using advanced computing tools.
Apart from that, the universities can develop associated research labs where the experts can design short-term courses and support the AI researchers who are really interested in AI. Note that it would be best to approach the field of computer literacy and novel tools such as AI in an environment different from that of traditional classrooms.
What advice would you give to Indian students working or aspiring to ML?
There are many research opportunities out there. AI is in the early stages of development and is growing and evolving rapidly. It’s one of the most exciting areas I’ve worked in. The job prospects are open, as is the competition. We are on a path where we still have a long way to go and we need more people who are interested in continuing the path. The question is, is it something you want to do? If so, I would suggest that you start by learning about math, probability, calculus, and statistics. Then get a quick guide on what AI research is all about, try small projects that closely match your current project, and learn how to solve the problem using ML or even simple data analysis using programming .
Above all, don’t learn programming by reading books, but by programming!
What books and resources would you recommend for ML aspirants?
There are many resources available. The beauty of the community is that most of the research is open source where you can see the codebase and even contribute to the project.
To get started, take free introductory mini courses, work on projects and slowly increase the challenges in the projects, learn Git and build a Git portfolio. Also, participate in ML contests and hackathons where you can cross-check different approaches of other participants. With all this, try out different methods and projects and stay curious.
