Meet Sreenithi Sridharan
Azure ML Graduate @ Udacity
United Arab Emirates
Sreenithi Sridharan put her passion and interest in AI to create solutions that can solve real-world problems. She is driven by passion rather than ambition and likes to live in the present without worrying about the past or the future.
Sreenithi has vital interests in Artificial Intelligence and Robotics and believes that these technologies can make the world a better place when put to use in the right way.
What inspired you to pursue a career in AI?
More than inspiration, I believe it was a fascination that led me to the path of AI.
I had attended a Robotics workshop as a child where we had to program a simple LEGO Mindstorms robot to follow specific paths. At that time, the joy of looking at the robot perform the actions you command was unmatchable.
I believe it was that fascination that's what made me search for a Robotics based course for my undergraduate studies though I didn't know anything about what AI or ML is about then.
Though I couldn't find a course exactly matching my interest in Robotics in the place I live in, I took up an alternative Computer Science that matched my interests at that time.
After I entered University, through the help of my professors and the subjects in my course, I slowly discovered more details of what AI and ML are. I gradually found myself interested in Computer Vision and its applications as it also has applications related to Robots and Human-Robot Interaction.
As a person, I also like to introspect on the way I think, understand things around me, and make decisions.
I found that many aspects of AI and Robotics do precisely this. While we try to think of ways in which we want a program or a robot to understand the world, it also helps to understand better how we as humans perceive and understand the world around us.
Hence I believe that the nature of AI itself inspired me to pursue AI as both my specialization during university and my career path forward.
Any AI project you want to share with us?
One of my recent projects was about training an image classifier that can classify Chest X-Ray images as whether they are affected by Covid-19, Viral Pneumonia, or Normal (affected by neither).
I took up this project as my Capstone Project for my ML Engineer with the Microsoft Azure Nanodegree program at Udacity.
I had the chance to choose my dataset and project topic for the Capstone Project and hence wanted to use this opportunity to develop something useful for the community.
The applications of Computer Vision and AI, in general, are very vast in any field, and I believe adopting AI in medicine can assist doctors in better diagnosing diseases.
Covid-19 has undoubtedly become an essential part of the medical field and the days coming forward in the current situation and effectively diagnosing the disease has become all the more important.
Hence, motivated by these thoughts and my interest in Computer Vision, I decided to create an image classifier that can distinguish between Normal, Viral Pneumonia, and Covid-19 affected Chest X-rays.
I have seen that Convolutional Neural Networks (CNN) provide perfect accuracy as they can effectively compress when it comes to image classifiers.
And learn features of an image through the different Convolutional filters following a linear idea flattening approach to train other classifiers and neural networks.
The effectively compress and learn features of an image through the different Convolutional filters rather than following a linear idea flattening approach to train other classifiers and neural networks.
This project also allowed me to recheck this as I had the chance to train and compare the performance of different ML models using the Python Azure ML SDK and Automated ML feature of Azureart of the Nanodegree.
The CNN Classifier gave me an incredible accuracy of approximately 97%, which I felt was an excellent result for the problem I was trying to solve.
I am glad that I got to work on such a project that was an example to show how AI can be put to good use by the people around and in the medical field at such crucial times.
Where could Deep Learning have a significant impact? And Why?
In general, I believe that Deep Learning and AI have a wide variety of applications in many industries, and hence I cannot pinpoint any particular sector.
Based on data and models, Deep Learning does classify into different Computer Vision, Natural Language Processing (NLP), etc. And in that perspective, one industry where I see a lot of scope for both Computer Vision and NLP is the field of medicine.
A popular misconception about AI is that it may replace a lot of jobs. Still, as far as the field of medicine is concerned, I believe that AI and Deep Learning principles like Computer Vision, NLP, etc., can assist doctors in making better diagnoses and predictions rather than replacing them.
In the book Deep Medicine by Eric Topol, the author talks about how the adequate time significantly reduced a doctor can give to the patient as doctors spend most of the time inspecting the patient’s reports to understand the patient's reports diagnosis from it. Hence, hardly any time is left out for the doctor to converse with the patient.
But Computer Vision and NLP algorithms can greatly assist doctors in summarizing and finding essential diagnoses from medical reports and scans in a much quicker way. Hence, this might help doctors spend more time conversing with the patients to understand their problems better.
In addition, as medicine is a very vast field, I believe that doctors too perform diagnoses based on the comparison of different symptoms, and their accuracy of diagnosing the right disease may vary with experience.
In such a case, AI and Deep Learning models trained over much more critical data than a single doctor might have seen during his experience can assist doctors in diagnosing the right disease much faster.
What tools do you like to use in your work and research?
There are two main components required for the successful completion of any ML or DL project:
The first and most important is the ‘Data.’ The availability of the data is what defines the type and outcome of an ML project.
Second is the ‘ML algorithm,’ which needs to be trained on the data to get a helpful Model.
I generally try to find openly available datasets from places like Kaggle and UCI Machine Learning Repository for the data.
When it comes to the ML Algorithm or everything related to the code, I believe the Python programming language is most useful and convenient.
From data cleaning and processing to building, training, and testing ML models, Python has its AI and ML libraries to do them all.
The most common libraries that I use in general are libraries like:
OpenCV, Scikit-Image, and Pillow for processing Image Datasets
Scikit-Learn for training general ML algorithms like Classifiers and Regressors
Tensorflow and Keras for Deep Learning algorithms
How do you see the ethical concerns of AI?
Well, I have heard of many different ethical concerns regarding AI. Still, the most common ones I have heard of are, first, the bias involved in AI algorithms and their results, and secondly, the concerns of people losing jobs due to the intervention of AI.
As a person with comparatively less experience in this field, I cannot thoroughly comment on the bias involved in AI. Still, as far as I have seen, an AI model is only as good or bad as the data it does train on, and hence the bias is also something that stems from the data used.
The bias that may be coming in is that the AI is learning from incomplete data and discrimination.
Already some awareness of this in AI developers, developers are encouraged to pre-process the data to eliminate bias.
In addition, through my Microsoft Azure Course on Udacity, I have also seen that Microsoft Azure has tools for building Responsible ML that can check and remove any inherent bias from the trained models.
Hence I believe that bias induced by the ML algorithms can be eliminated to a certain extent by proper preprocessing and checking of the data and models.
And coming to the replacement of jobs, I see AI as a new technology getting embraced by this world now and can improve and bring about good changes in the world.
In that perspective, just as I mentioned in the medical field, I think that many AI applications only involve assisting humans in doing their job better rather than completely replacing human experts in the area.
Through automation, I agree that the need for human workers to perform specific repetitive jobs will reduce. Still, at the same time, AI can create many more new and different job opportunities than the ones it removes.
And this is the case not just with AI but with any other new technology or industrial change that has happened over the years.
Just as how the advent of computers first removed the need for people managing records and jobs manually, in the long run, it enabled more people to learn about how to work with and control computers, which has now changed the world to an entirely new level.
Similarly, with AI, I believe that it will only create newer job opportunities and enable people to learn more and adapt to the incoming change.
Hence to conclude, what I would like to say is that ‘Change is the only constant in this world.’ If it’s AI in this decade or century, it’s going to be something else in the future, and we either have the option of embracing AI and the changes that come with it for a better lot or fear the consequences and stay the way we are.