Interview with Subash Gandyer, Senior AI Lead, ConversationHEALTH Inc

ConversationalHealth is a company that provides a Conversational AI platform for global pharmaceutical and life science companies.

We thank Subash Gandyer for participating in the Conversational AI Interview Series and sharing several insights, including:

  1. His responsibilities as a Senior AI Lead.
  2. How conversational AI is leveraged in the healthcare sector.
  3. Resources that helped him in his data science career and what drives and inspires him to work in data science.

1. What is your background and how did you get involved with conversationHEALTH?

I have a Masters in Computer Vision and Image Processing with more than 10 years of experience in Machine Learning and Deep Learning respectively. I started my career as a tutor helping college students learn Math, Computer Science and Machine Learning. I loved teaching and I became an Assistant Professor in the School of Engineering at leading universities in India.

Then, I got involved in a couple of ML startups as an AI/ML Consultant and slowly transitioned to a full-time practitioner. I took various roles such as Machine Learning Developer, Deep Learning Engineer, and Data Scientist. I was building Recommendation Engines, Product Sentiment Analyzers, Brain Computer Interfacing Applications to name a few.

After moving to Canada a couple of years ago, I started working as a Data Scientist in the healthcare space building NLP parsers for doctors encoding prescriptions, discharge notes and clinical text. This past December, I joined conversationHEALTH because of three important elements that I love most: AI, healthcare, and chatbots. The conversationHEALTH SaaS platform enables exactly this – intelligent chatbots for healthcare companies to bring their business quickly and efficiently into the conversational age.

2. What are your responsibilities as a Senior AI Lead at conversationHEALTH?

My responsibilities with conversationHEALTH spans these five-folds. First, my priority is to increase the response rate of any queries asked of a conversationHEALTH bot by analyzing fallbacks and creating scalable NLU modules.

Secondly, my major focus is to oversee our client bots to ensure the highest level of accuracy possible by developing and incorporating State-Of-The-Art algorithms into our solutions.

Thirdly, I automate manually-intense internal processes of conversation writing, thereby reducing the time-to-market from a couple of weeks to a few days. I also investigate, study, and experiment with the existing literature for a specific problem identifying the best possible solution. Last but not least, I mentor and transfer learnings to our NLU team on the best practices available for developing the best solution to the problem at hand.

3. Who are conversationHEALTH’s customers and how do you use AI and data science to create value for them?

conversationHEALTH’s customers are major global pharmaceutical and life science companies. Some of our customers include Bayer, Merck, Lilly, AstraZeneca, among others. Our clients span across 7 countries in 3 continents.

conversationHEALTH is the cloud-based conversational AI platform for the global Life Science Industry. Its Conversation Management Platform powers Sales & Marketing, Medical Affairs and Clinical Trial solutions that help these companies engage healthcare professionals, patients and consumers, 24/7/365 through text, voice and digital humans.

We launched 2.5 years ago and have already partnered with 16 global pharma and life science companies.

4. Which processes are you helping your customers automate with bots & Conversational AI?

Conversational AI allows our customers to scale conversations so their customers, whether that be patients or healthcare professionals, are engaged 24/7 by personalized and on-demand interactions. Implementation of our virtual assistants have the ability to leverage a common NLP and Machine learning model across the business that drives exceptional user experience and accuracy, as well as provide cost reduction. For example, the chatbot can decrease the number of live agents required at a call center while increasing frequency and quality of engagement with customers.

Our patient-centric bots can reach patients on their own terms, answering their health questions and providing support 24/7. In addition, our bots can initiate critical conversations with patients at the right time in their healthcare journey by delivering information that is both on message and empathetic. Patients can now ask critical emotional questions that they have not felt comfortable asking to date. We recently created a chatbot for the Asthma and Allergy Foundation of America for personalized support for patients, consumers, and caregivers, as well as for insurance navigation.

We have also developed an HCP Assistant for healthcare professionals such as doctors, nurses, and paramedics to gain knowledge of specific drug-related products and to receive the latest clinical updates. This bot provides information about the Indications, Side Effects, Drug Interactions, Dosing, Efficacy, and other related medical queries.

We have other interesting chatbots in the prototyping/development phases.

6. How important is the domain knowledge of the business/industry you’re in as a data scientist, and how did you acquire it?

As a matter of fact, for any data scientist in any domain, having a basic knowledge of the domain is a must. If you understand the domain, the solutions you come up with for a problem would be very intuitive and most of the time, it would be the best solution possible. Having an expertise level of understanding of the domain is nice to have. Particularly, for the healthcare space, it is absolutely essential to have a basic understanding of the ecosystem. At conversationHEALTH, we have a cross-functional team of data scientists and NLU Specialists working in tandem with Medical professionals on our Taxonomy team to ensure medical understanding, nuance and compliance is met.

I acquired this domain knowledge over a number of years of watching my parents, both of whom are Physicians, taking care of their patients. Although the Indian medical ecosystem is different from the Canadian healthcare system, the fundamentals remain the same. The stakeholders remain the same including; Patients, HCPs, Hospitals, Insurance Companies, Government and so on. Only the inner interactions within these stakeholders change a little in both of these ecosystems. When I moved to Canada, it took me a whole year to understand the inner workings of this entire ecosystem. conversationHEALTH has now expanded into a global company, supporting businesses in North America, Europe, and Asia and it is absolutely essential to have domain expertise for a data scientist to build global solutions spanning different ecosystems.

7. What are the biggest challenges that you are currently facing at conversationHEALTH?

One of the major challenges I am facing from both a business and technology perspective is human level accuracy of our chatbots. As we all know, we are nowhere near that level of accuracy because NLU is such a difficult problem to crack. At conversationHEALTH, we are as close as possible to the existing State-Of-The-Art chatbots in the world. Time, Data and Computing are the answers to achieving this goal in the future.

Secondly, scaling of trained meaningful conversations from one domain to a completely new domain is a challenge. For example, if we build a bot for diabetes, the conversations are not fully transferable to a chatbot that deals with mental illness. The intents, entities, and other intricacies are very different including the way each patient or caregiver would interact with the bot.

conversationHEALTH is solving these challenges with an efficient, robust framework including multi-level cross-validations, fallback detectors, taxonomy-oriented solutions, and many more.

8. What are conversationHEALTH’s biggest achievements in the last 12 months?

Some of the major achievements for conversationHEALTH in the last 12 months are:

16 global life science clients, working across lines of business, with solutions that are now scaling globally.

We recently surpassed a team of 50 with cross-functional expertise in healthcare, technology, and conversational design. We continue to grow rapidly to meet the growing needs of healthcare.

95%+ of questions are successfully managed by our bots in the market.

9. Which Conversational AI-related technology trend do you think will have the biggest impact in your industry in the coming years?

Digital Humans with Persona, able to cater to the needs of Patients and Caregivers.

Voice Assistants with Empathy, enabling people to achieve their goals.

Assistants that support individuals along personalized journeys.

Assistants that communicate with a patient depending on their medical literacy level, and much more.

10. As an experienced data scientist, please share with the readers some of the best resources that helped you grow as a data scientist (it can be books, courses, blogs, etc.)


Machine Learning – Tom Mitchell

Deep Learning – Ian Goodfellow

Pattern Recognition and Machine Learning – Christopher Bishop

The Elements of Statistical Learning – Trevor, Robert and Jerome

Deep Learning with Python – Francois Chollet


Machine Learning – Andrew Ng (Coursera)

Deep Learning – Andrew Ng (Coursera)

Stanford NLP lecture video series

Blogs / Influencers

Andrej Karpathy

Lex Friedman

Machine Learnings

Google AI Blog

Berkeley Artificial Intelligence Research (BAIR) Blog

Open Source Github Repositories

Google AI Research

Microsoft AI Research

Facebook AI Research

Hugging Face

11. What inspires you about working with Data Science and Conversational AI?

As I have already mentioned, I was working with Computer Vision domain analyzing images as well as processing them and producing models. Now, in the realm of Text domain, I find this Unstructured Textual domain very challenging compared to image processing problems. This inspires me to push the boundaries in finding solutions for hard text problems. Saying that, the field of NLP has improved so much over the last few years.

Although we have made small successes in breaking benchmarks in applications like NER, POS, Question Answering, Classification with SoTA algorithms and techniques, we are nowhere close to the human level of accuracy in holding meaningful, interesting, and empathetic conversations. These prevailing challenges inspire me the most to find optimal solutions.

About The Author

Scroll to Top
Share via
Copy link
Powered by Social Snap