As major pharma companies in India continue to grapple with GMP compliance issues, a startup combining deep learning, AI tech, and NLP models based on artificial neural networks is helping subscribers be better prepared for inspections, learn from the US FDA trends and improve their GMP processes continuously based on those learnings. With the official launch planned for February 20, Bhavish Agarwal, founder, FDAlytics.com tells Viveka Roychowdhury how his company’s algorithms can augment the analysis of pharma regulatory affairs personnel by providing them up-to-date and readily available data points
Bhavish, thanks for talking to Express Pharma. Tell us something about your previous life before you turned entrepreneur.
Thank you. It is my pleasure to speak to you. I have been working in the field of Artificial Intelligence (AI) for the last 10 years in companies such as Microsoft, Zlemma and Hired.com. I met my co-founder Chakri Uddaraju at Hired. He was working at Facebook building the Marketplace before we decided to start Fdalytics.com.
How did you find yourself starting an ML/AI startup for pharma companies in India after a career in Microsoft, etc?
It has been an interesting journey for us. We wanted to use our experience of working at big tech companies in cutting edge technologies in the field of healthcare and pharmaceuticals. During our interactions with multiple pharma players, we realised that GMP compliance remains one of the biggest and continuous challenges.
We identified the US FDA GMP regulation and compliance as the problem and started thinking about the various ways we can use our experience in deep learning and AI tech to solve this problem. We found two opportunities: First in terms of India’s growing position in the API and finished dosage exports to the US. By being in India, we are getting easy access to an industry which is looking out for innovative solutions to minimise or eliminate their compliance-related risks.
Secondly, once we started capturing the US FDA data, we immediately understood how ML/AL can become a big differentiator and create value for our customers.
In summary, strict monitoring of GMP laid down by the US FDA and vast amount of data where ML/AL can be used made it a no brainer for us to start an ML/AI company for pharma companies in India.
What is the basic rationale for FDAlytics.com? What are the features of FDAlytics.com like inspector analysis, etc?
The aim of Fdalytics is to provide data and insight for GMP compliance to global pharma and medical devices companies. We want our customers to be better prepared for inspections, learn from the US FDA trends and improve their GMP processes continuously based on those learnings.
At a high level, Fdalytics features can be described in three categories: Data, Search and Intelligence.
As of today, the FDA inspection related data is either not available, scarcely available or if available, it exists in disjointed and siloed formats. It is virtually impossible to gather or combine these data points and gain insights from it.
If a pharma company wants to ask insightful questions like – what is the most frequently cited CFR violation last year? What area is a specific inspector more likely to focus on based on his or her behaviour? What are the global inspection trends based on all inspections? This is almost impossible to find out yet this is very useful for a company to go through the inspection.
Let me give first give you an overview of the data that we have before I tell you how we use ML/ AI to leverage these data points.
In Fdalytics we have gathered data points for all the inspections conducted by the US FDA since the year 2000. For each of these inspections, we have collected whether a Form 483 was issued, whether the inspection resulted in a VAI, OAI or NAI, how long that inspection lasted and who were the inspectors who conducted these inspections. We also have tens of thousands of 483s, EIRs, responses and Warning Letters.
Alternatively, the platform can also tell you the track record of each of these US FDA inspectors. For example, what inspections they conducted, how many 483s they have issued etc. Since we have these track records our users can also gain insight on each of these inspectors. Additionally, our users can use the advanced search features for each inspector to drill down to a single observation made by the inspector during an inspection.
But this is just the surface. We are using Natural Language Processing (NLP) models based on artificial neural network. These models operate upon tens of thousands of Form 483, EIR, warning letters etc to pinpoint the specific areas that are the most relevant to an individual customer.
We are creating features where based on a company’s profile and track record, Fdalytics algorithm can make personalised recommendations.
Pharma companies in India need to analyse the technical details of US FDA inspection reports, be they 483s or EIRs, on a daily basis. How does the ML/AI tool score better than regulatory affairs personnel in pharma companies who have had years of experience within the sector?
Yes, you are right. Regulatory affairs personnel need to go through a lot of 483s in order to learn from US FDA observations. However, there are many challenges to it – availability of data being the number one. With FDalytics, we solve this basic but foundational problem by providing the data in a click of a button. You no longer have to ask or call your friends or colleagues to get second hand or incomplete information or pay an enormous amount of money.
As you said, most of these personnel have years of experience within the sector. With Fdalytics, we aim to augment their analysis by providing them up-to-date and readily available data points. Since Fdalytics is a technology platform, we can scale it up which is very difficult to achieve without technology.
We are very confident that Fdalytics will become an example where AI technology and professional experience come together to create a far more efficient environment for GMP compliance. In short, FDAlytics is an intelligent assistant platform for pharma companies and regulatory / compliance leaders to help them in GXP.
Give us an example to illustrate. Suppose a company needs to look for all data integrity related 483s dealing with injectables.
This is an interesting example. Before Fdalytics, this problem would have taken a lot of manual efforts and resources, assuming someone has enough 483s. Then too, the chances of missing an important observation would have been high.
In the Fdalytics platform, our user simply has to search for the keyword ‘Data Integrity’ and chose the category ‘Injectables’ and our algorithm would go through almost 30-40 thousand individual observations and find the relevant observations.
You can even refine your search so that you can find specific issues by a particular inspector. For example, you can find all ‘computer system validation’ issues cited by say ‘Thomas Arista’ etc. Not only that, the search result can be filtered where you can filter those observations that resulted in a warning letter. All of this happens on Fdalytics in few milliseconds.
Our customers love the simplicity and the ease of use of the platform. But in the background, our AI algorithms are hard at work. For example, if some searches for ‘Data Integrity’ the algorithm also understands that phrases such as ‘electronic data’ or ‘reviewed for accuracy’ are closely related to data integrity.
One more example of a solution that we are providing is searching for observations by CFR citations. Before Fdalytics, it was virtually impossible to search for issues related to say, 21 CFR part 11. Our powerful and accurate algorithms are capable of mapping all the individual observations with their corresponding CFR code. This is a powerful feature, and few of our customers are already using to update their training programmes and SOPs based on these CFR examples.
What is your team size right now and how are you funding this venture?
We are a team of 10 passionate and energetic people, majority of whom are engineers and ML researchers. We have also onboarded pharma industry experts with vast experience and unparalleled domain expertise. We are completely bootstrapped and have already started generated revenues.
With the official launch planned on February 20, what has been the response during the past few months in the beta launch stage?
We launched a few months back in the stealth mode and worked with select customers. The response has been phenomenal and provided testimony to the fact that there is a real demand for technology tools to help companies navigate the complex world of GMP compliance and regulation.
The feedback was very encouraging and gave us the confidence to launch Fdalytics officially. We also realised that the pharma manufacturers are very tech-savvy and are fast adopters of the latest technologies.
How many companies have signed up? Are they all pharma companies or some are consulting companies as well?
We have close to a dozen companies both in India and overseas who are already are using the Fdalytics platform. We also have few pharma consultants who are using the tools to keep abreast of the developments in US FDA conducted inspections. We are actively receiving demands and interests from new customers.
Give us an idea of the subscription schemes on offer.
First of all, anyone can sign up to Fdalytics.com and start using certain search features for free. Fdalytics comes with a lot of flexible and affordable pricing option. We understand that each pharma company is different and will have different requirements.
Depending on the requirements, one can purchase Form 483 or an inspector’s report. We also offer unlimited access to Fdalytics platform as a yearly subscription by paying
a one-time yearly fee.
Our subscription customers get unlimited Form 483s, EIR etc download, access to all the inspector profiles and access to all the advanced features search as advanced search and recommendation. We also build custom machine learning models for our subscription customers that give specific insights for their companies.
Who are your competitors in this space and what is your USP?
Fdalytics is the only pharma GMP compliance platform that uses AI/ML and operates on US FDA data. This is the first time that advanced NLP and neural networks are being applied in this space.
There are few websites which sell Form 483s etc but the customers have to pay exorbitant amounts to get just one such document. Our customers not only get these documents at a much cheaper rate but are able to leverage the power of big data and AI technology to get actionable insights. In essence, our USP is our data and technology.