A Knowledge-Provoking Interaction with the AI TEAM of SMI

   The PSYEC-AI Time Journal AI Club had a very healthy knowledge-sharing interaction with the AI crew of Sri Mookambika Infosolutions (SMI), Madurai. We had so many questions with us and we greatly utilized the opportunity to get a clear view of AI and its possible applications along with the challenges they are facing. I believe this conversation will help the readers, AI Entrepreneurs and Students to have own Start-ups.

SMI marching towards its goal of setting values in Emerging technology with Eminent Members as follows,

  • Mr M.P.Selvaganesh –  the Chief Executive Officer,                                         
  • Vice Presidents 
    • Mr S. Prasanna for Technology                                             
    • Mr S. Shanmugam for Machine Learning                                           
  • Data Scientists 
    • Mr M. Lakumi Narayanan, a Data Architect                       
    • Mr L.Muthuvel, a Data Scientist.

We (PSYEC AI Club) shared our experiences with SMI. We shared our knowledge in Artificial Intelligence and writing articles about both interviewing and content marketing. Thus our discussion begins.

During Discussion

1. How do you apply AI in SMI?

AI is applied in SMI for various RnD activities and also to address specific customer requirements. We started with an usecase for automating a Quality control process, thereby slowing extending into various other areas such as Document processing Automation for scanned images (Using OCR), Predictive Analytics, Industrial Automation, ANPR, Behavioural analytics – Human, Traffic, Computer vision related video streaming analytics etc. We do have a strong Data scientist team to take care various use cases and explore algorithms in detail, whenever the situation demand it.

2. Tell about SMI and your products

We have above 460 workers in SMI and have multiple branches in Madurai, Chennai & Rajapalayam. SMI is a community (more than a corporate), a community that grows by empowering the people within. We have clients all over the world covering various domains. We are strong in Software Services for the last 12-13 years, but we started gaining a lot of traction and expertise in Products, SAAS models and Analytics as well in recent days. One of our AI solution is the Document Automation process which extracts & converts images (scanned fax images) into meaningful Electronic data automatically. This automated process helped in reducing a huge number of manual effort, by atleast 50% (200 people to well below 100 in one such instance). There by saving the cost associated with it along with avoiding to take care of such cases in various domains and markets across the globe. We have ChatBots for Healthcare & Other CRM application, which picks up the appropriate answer for us & interacts with the end users. Building a Chatbots requires a large collection of data in which data are mostly content/domain specific. In addition, we do Optical Character Recognition solution, Predictive Analytics Solution and so on.

3. Challenges faced in this domain

One of the prominent challenge is the Data collection, takes six months to one year. It is the biggest challenge because sometimes we need to pay for the data or otherwise we need to prepare the data manually (synthetic data). Speaking of sensitive domains such as Healthcare, we have to be very cautious about the data, as we have to agree & follow with some compliances.

4. Why AI and ML not reached in remote areas?

There are some difficulties for AI to implement in remote places. We need resources for training and utilization. We need terabytes of Memory and Internet facility to access the cloud for enormous storage. Such facilities are not available in villages yet. So it takes time to carry AI to remote places.

5. Give brief information about tools and programming languages for AI and ML

There are different tools for various platforms. Python is used for ML applications a lot of libraries are built in. Java is faster in performance comparing to Python. R is suitable for Statistical functions and Hadoop for big data analysis.


Python for ML

Java is Fast in Performance

R for Statistics

Hadoop for BigData

6. How to get started with AI projects?

First, analyse the problem and try to get the solution with a simple programming language. When human intelligence about a logic to solve a problem struggles, then AI comes to solve it. Programming languages are used based on the use cases. Otherwise in a nutshell, to start with AI projects, 1. Understand the problem statement, 2. Collect & Analyse the data, 3. Research & Choose or tailormade the perfect algorithm which addresses the problem and the regular game goes on.

7. What is Data science and how to become a data scientist?

Data science is the collection of proper data so that it is fed to the machine to train. Understanding Math fundamentals are the base for becoming a Data Scientist. Learn how the human brain take decisions based upon the parameters. Look into the data, observe and get the fundamental working behind that. Don’t think if it happens it is. Try to figure out what is happening behind the screen? Remember nothing comes at random so reason the cause. In SMI we have technically fit data scientists – More than hiring people, we groom them.

8. We want to know about TensorFlow

Tensorflow is just an open source machine learning library/collection of libraries which has built-in algorithms & utilities to support things around it. Keras is developed on the top of the tensorflow. There are more packages present nowadays to support developers which enable drag and drop features mode of development.

9. How do you handle the missing data?

Data handling is somewhat a tedious process. One of the complexities is handling the fake and missed data. So we need to do Data cleansing and filtering. There some unwanted data gets removed by our data scientists. Without completing the process we cannot reach further data. When it comes to missing data, based on the nature of the other data/metadata present, we do synthesize the information depending on the situation.

10. Share the disadvantages of AI as it replaces many employees

It’s not a disadvantage when Automation (not only AI) replaces human workforce. The same human workforce can be leveraged for doing something else better than that. We need to evolve as how the machines do these days. Also still we need men to prepare the data, framing, filtering, training and maintain it. We need to have someone to train the model so that it’ll be a beneficial one. We need employees to make AI possible.

11. Vision and mission of SMI

Our Vision

 To be the Technologically advanced & most admired Organization providing high quality and cost effective services with a Long-term client relationship strategy.

Our Mission

Utilizing relevant Technology and the specified knowledge to provide unique value and high quality solutions to our s, serving with integrity and fairness.

Opinions expressed by AI Time Journal contributors are their own.

About Ebin Rose Christina

Editorial Staff Intern Student @ Pandian Saraswathi Yadav Engineering College

View all posts by Ebin Rose Christina →