Artificial Intelligence and Machine Learning are rapidly advancing. Upskilling in these fields can help you power ahead your career. Read further to learn more about Krishna’s journey with Great Learning’s PGP Artificial Intelligence and Machine Learning Course in his own words.
I am an Electronics & Communication Engineer with 21+ years of Experience in Software Development, Platform Engineering, DevOps, Product Management, with the majority of my time spent in the Financial sector.
Before I took up this program, I was doing a full-time job as a Senior Project Manager in Fidelity Investments, Bangalore. I was heading the DevOps India team. To pursue AIML, I very clearly knew beforehand that I could not dedicate time to learning with a hectic weekday and weekend schedule in the office. Hence I quit Fidelity Investments in Aug-2019 to focus on AIML learning. Of course, this is the most bizarre & unusual thing anybody with a full-fledged career would do in the Indian context!!
Well, I was watching the yearly Google conference that Sundar Pitchai addressed in 2018, where Sundar showcased a collaboration between Google & Sankara eye care(India) case study video on eye care based on AI. I was really impressed with what AI could do in detecting the eye conditions of patients, along with predicting various other ailments that the patients were actually facing but were unaware of !. I was really inspired and impressed by this talk and started exploring more on the beneficial aspects of AI to help humanity… After this, I started trusting in AI, and I started imagining the kind of use cases that AI / ML could potentially solve. This was my starting point to learn AI / ML.
To talk about my capstone project – If I give an x-ray image of the film, can the AI system predict or infer the type of disease/condition or pathology in medical terms ?. In the current Medical system, this question is answered by an expert qualified Radiologist who answers this question as part of the X-Ray report. This report is studied by the Qualified doctor or Pulmonologist ( Lung specialist ) to infer the lung condition before starting the treatment. This capstone does not try to replace the Radiologist rather augments or collaborates with the Radiologist in inferring the lung pathology. Over the period of time, as the Radiologist establishes the trust of the system, the Radiologist can act as an approver of the report. This has immense applications in telemedicine scenarios & in remote areas where expert Radiologists are very sparse. This system will significantly reduce the existing load and scale their expertise.
When we started the capstone in Apr-May 2020, our team had an idea to do a project in the Medical field. We selected a project to apply AI on X-Ray or CT scan images to contribute to the Medical field. After around a week of selecting the project, our Professor, Dr. Narayan, came up with the idea of doing something worthwhile in the area of COVID since it was the early days of Pandemic. He wanted our team to contribute AI research in this field and shared with us a few research papers related to this field from the University of Waterloo, Canada. This was the starting point. Our team was excited to apply AI in this emerging area and was curious to see the findings.
There were many challenges faced in the project, namely – a) Time of engineers since most of them had full-time jobs, b) Compute challenges – We used Google collab free version as the platform. AI is very heavy on computers, and we had to work around the restrictions. c) Choice of algorithm and directions – Our mentor was very helpful and listened to our progress and roadblocks, and gave appropriate directions to experiment. This helped in confirming the path that we had taken and gave us more convictions to do new experiments and progress the project/research. d) Data deficiency of COVID x rays was a major roadblock – We discovered new trusted COVID data sources over a period of 7-8 months by going through the research papers. We refined the models at least three times after getting new data sources. This effort really paid off in the form of better results for existing models. Since our team started with a very lean data set for COVID, we devised methods in our algorithms to train with imbalanced data sets. These strategies help to fix these major roadblocks.
I think mentorship played a very important role in succeeding in this project. I would attribute great mentorship to both Dr. Narayan and Mr. Abdul. This helped in confirming the path that we had taken and gave us more convictions to do new experiments and progress the project/research. Many times we learned new concepts in AI to aid the new experiments and see the light of the day.
This project was purely research-based with no organizational goals. In this project, we were able to highlight the unexplored algorithms that can make a big difference in detecting lung pathologies very accurately with less compute resources ( DarkNet-53). The algorithm can accurately detect pathology even with very few samples for the class of interest (Ex- COVID). Also, this algorithm can be trained on either X-Ray or CT-scan images with potentially. This paper also gives a comparative analysis of various Transfer Learning algorithms and how the DarkNet-53 outperforms other algorithms with the same dataset. This research paper was accepted at the CD-MAKE International conference in Austria-Graz. The team presented this paper on 18-Aug-2021 virtually and received a lot of accolades, especially on the applications. Many attendees were curious to know if this is being productionalized by some startup or company? The paper was published by springer Aug-2021 in this journal – https://link.springer.com/book/10.1007/978-3-030-84060-0
Our capstone team is fairly senior with an average experience of more than 15+ years in the Software Industry. This project has highlighted the experience and research capabilities of our team. After this paper got recognition in an international conference, we have received a lot of recognition from our Great Learning Professors, fellow classmates, family, friends & my fellow colleagues in my last company. This has given more thrust and boost in self-confidence in exploring new areas, especially in the application of AI to new domains.
Age & Experience are numbers that increase the weight of ego, risk of failure and inhibit one’s passion, curiosity, exploration & experimentation, be it in personal life or professional life. I feel shattering these glass walls will help in unlocking one’s potential and exploring exciting unchartered territories. b) This is another new learning as I joined a startup doing ML working on structured data: While we learn all the fundamentals and happening algorithms/models in Great Learning, Domain context, Data & Communication is as important or more important in putting context to all the data and algorithms. This will put things in perspective in providing solutions to business problems. This will certainly add more fuel and growth to respective AI/ML careers. c) Lastly, attitude towards Business problems, Domain to your fellow team members, and being result-focused covers the last mile. I would like to close my advice with these little pieces of wisdom that I gained from my learning tenure in Great Learning and the new Data Science startup I just began working for.
