Listen to the content...
Voiced by Amazon Polly

In this interview with Ram Subramanian, chief technical officer at PerceptiMed, we dive into what artificial intelligence (AI) really is, what separates robust AI from gimmicks, and how pharmacies can be deploying AI to solve everyday tasks in the workflow. Plus learn more about how PerceptiMed is using AI in its products.

ComputerTalk: Ram, we hear a lot of buzz about artificial intelligence or AI. What is AI?

Ram Subramanian PerceptiMedRam Subramanian:In simple terms, we can think of AI — which is also referred to as machine learning — as a machine that takes data as an input and, based on a very large reference data set, learns something meaningful and produces decisions. For example, if the machine is fed images of drugs, it can identify what drug it is. But a considerable amount of work has to be carried out to make the system robust and repeatable. In fact the amount of this work put in to make an AI algorithm or task robust is what differentiates very useful AI or machine learning enabled tasks from ones that are gimmicks.

How can AI be relevant and impactful for pharmacy?

Subramanian: Every day pharmacists and pharmacy staff have to make hundreds of decisions based on questions such as:

  • What medication is this pill?
  • Did I count the pills in this fill correctly?
  • Should I order more inventory for the flu season?
  • Will this prescription be picked up if I reach out to this patient to remind them, or do I just return the prescription to stock?
  • Do I need extra staff during certain days or months?

These are questions that can be answered efficiently by an intelligent machine.

What AI products are you working on at PerceptiMed?

Subramanian: At PerceptiMed we are focused on solving the issues related to medication patient safety, chain of custody, and medication adherence using AI and automation. Currently, we have one commercial product, scripClip. It is a will-call solution. We have two other products, identRx and MedPass, which both heavily rely on AI for their functioning. These devices identify medications during the medication filling process. The devices capture images of pills being filled and use AI to extract visual features to identify each pill. We have spent years developing our AI algorithms to achieve near 100% accuracy on thousands medications.


The scripClip will-call system helps pharmacy staff quickly and efficiently locate the right bag for the right customer, every time. With analytics, the pharmacy has access to different metrics regarding sales, and can leverage the data to make informed decisions.


identRx is the industry’s first counting system able to accurately read visual features on every pill as it is dispensed.


MedPass was specifically designed to eliminate the margin of human error in administration of pills in long-term care facilities.

This is where our product, identRx, comes into the picture. identRx is an intelligent machine that triple counts and verifies each drug in real time. Its verification accuracy is 99.9%. It can save a lot of time and money for a pharmacy so you can spend more time with the patient.

What hurdles to adoption do you see?

Subramanian: Adopting new technology is not always easy. Educating customers about the benefits of AI in pharmacy is crucial for product adoption. In many cases products and services claim to use AI analytics but merely use simple statistics and good visual graphics to show patterns. Part of the problem is that many people are unsure about the applications that AI can solve. So we have an added need to prove to the users the efficacy of the AI system. At PerceptiMed we have designed the products with simple and clean user interfaces. We focus on reducing the human effort required for repetitive tasks so that they are easy to perform. This will also allow the staff to focus more on quality patient care.

Do you see this technology fitting in with the current workflow in the pharmacy?

Subramanian: Absolutely — think of the workflows in the pharmacy today.

  • The staff counts pills by hand.
  • Pharmacists verify the fill.
  • Prescriptions sit in will-call and are returned to stock in 14 days.

All of these are, often, currently manual efforts that take the staff away from interacting with patients. If we can reduce these manually intensive tasks, staff can then operate at the top of their profession, with counseling, immunizations, etc. So for example, as we discussed earlier, the medication verification step and the product-filling step can be easily automated with a device that can not only accurately count the medication but can also verify the pills using AI as they are being counted. Considering that a good portion of a pharmacist’s day consists of verifying the medications being dispensed, this sort of device will definitely help free up a considerable amount of the pharmacist’s time to focus on more patient-centric care.

What are some recommended resources for people interested in learning more about AI?

Subramanian: These days AI resources are very abundant. The AI community has gone to great lengths to make the most popular AI tools very accessible to all. Some of the most prominent resources are geared toward an area of AI called deep learning. Stanford, MIT and others (Coursera etc.) have online courses. Obviously, the understanding of theoretical concepts are great, but is only half the story. Having the ability to develop a solution is the other half. For this purpose, the AI community has made most deep learning tools open source. The infrastructure to train or create a deep learning solution is easily accessible via many cloud-based systems (e.g., AWS, Google Cloud, etc.). Putting all this together, one can produce an AI-driven solution if the required quantities of data are made available. CT