Veera Raghavendra Chikka is a Senior Data Scientist at Ericsson Global India.

He is a natural language processing and machine/deep learning expert.

INDIAai interviewed Veera to get his perspective on AI.

From a software engineer to a senior data scientist at Ericsson Global India. How did this transformation occur?

After my graduation, I worked for a startup as a software engineer, where I contributed to different projects on social media analytics, Enterprise search engine, and Natural language processing. My work involved handling vast amounts of data using Hadoop, HBase and Lucene, which come under "Big Data analytics". Then, I moved to IIIT Hyderabad to pursue my PhD, where I saw the research focus moving from Big data to Machine Learning/AI. So the transition to ML/Data science was a logical step now.

What motivated you to continue your career as a research assistant?

Research Assistant is a role that you take as part of a PhD. When you ask for motivation, I think motivation is a feeling that helps you to do a small task at hand, and we cannot rely solely on this unreliable feeling for long-term goals. If you are obsessed with a goal, when the goal is challenging enough, and if you never give up, you will continue your journey no matter the hurdles that come your way.

What inspired your interest in artificial intelligence?

Everyone would get fascinated by any new technology that comes to market.

The capability of the technology will inspire you. I am inspired by the wide range of problems that can be solved by Artificial intelligence. AI is something that is here to stay for a long time.

What are the most widespread myths you would like to dispel as a long-time member of the AI and machine learning community?

For someone new to this field, I can point out two myths:

It takes a long time to learn AI/Data Science.

• AI has been there for more than five decades, even though the keywords we use to describe it may vary, such as machine learning, statistics and neural networks. To start a career in data science, you don't need to spend 1 to 2 years of coursework. Spend the first iteration of 2 months learning the basics of python, machine learning and neural networks. Pick a Kaggle project and have hands-on solving the assignment. Then, go for a second iteration by doing projects one each on Classification, Regression and Clustering. In my opinion, learning by doing is a better way to learn any technology. 

Everything is a black box: 

• Every algorithm in machine learning is backed by solid statistical theory. There is a new segment called Interpretable AI to understand a machine learning model using techniques such as LIME and SHAP.

Everyone acknowledges that machine learning and deep learning are innovative technologies. However, we frequently miss the new issues they present. One of them is the environmental impact of these rapid technologies. What do you think about that?

Any new technology, such as blockchain, and AI, despite many breakthrough achievements, impact the environment because of their resource-intensive nature. For example, AI, specifically Deep learning, demands high computing GPUs, 100s of GBs of memory and runs for hours on training. However, we can significantly reduce their impact by only using the most relevant data, adapting transfer learning and using model compression techniques. 

What factors should organizations consider when developing an AI roadmap?

Not all projects need AI. Problems come first, and technology comes later. Organizations should rather have answers to the below questions:

1. What is the customer's need?
2. How does the end user consume the outcome?
3. Identify the problems that need AI or machine learning.
4. Do we have enough relevant data to build an ML model? 

If we have clarity on these questions, we can solve any problem. 

If hiring, what skills do you anticipate from a fresher?

In addition to technical skills, I look for a positive attitude toward challenging problems and the ability to explain the solution in layman's terms. 

What advice do you have for those aspiring to pursue AI careers? What are the best paths to advancement?

Solve as many Kaggle projects as you can. In Kaggle, a problem statement is clearly defined, and the data is ready to use. Your task is only to do exploratory data analysis, model training and evaluation. In contrast, in industry, we define the problem, check if it can be solved using AI, define the scope of the problem, collect the relevant data, and transform the data for model building. Solving a wide variety of issues enhances your chances in job interviews. 

MLOps and cloud deployment would give you an edge in your profile for quick career growth.

Could you provide a list of notable publications and books on AI research?

I prefer reading research papers, articles, and blogs to be updated. 

After a point, following the latest trends would become difficult. So I also rely on research paper aggregators like paperswithcode.com and nlpprogress.com that highlight recent trends in ML and NLP research.