Recently, at Ryght we have had a surge of inbound interest from VCs who are convinced that #verticalSaaS is going to win in #generativeAI. Also, looking at earnings reports from companies like C3 AI whose CEO, the famed Thomas M. Siebel (who I had lunch with many years ago when he was starting the company), it seems that #EnterpriseAI is taking off.
Tom recently stated, "The enterprise market is on fire.... we are seeing overwhelming demand". At Ryght, we have chosen to live in the intersection of Enterprise AI and Vertical SaaS, making sure our technology is the very best in healthcare & life sciences and enterprise AI. Here's why:
Real Time Genomic Analysis in Oncology
My last company, Deep Lens, which I co-founded in 2017, took the concept of enterprise cloud-based AI and applied it to the healthcare and life sciences space. It paid to stay focused and go as deep as possible in solving expensive problems for customers. I like to call that "a problem with a checkbook attached." We did that in spades, and it worked well.
Without staying incredibly focused on very hard to solve problems for a narrow set of community oncology providers we never would have been able to match hard to find oncology patients with precision oncology trials from biotechs who were looking for the needle in a haystack - patients with specific genetic mutations and a specific type of cancer. We used natural language processing (NLP) to sift through millions of pathology and EMR reports from vendors like Flatiron Health CureMD Epic Cerner Corporation / Oracle , and genomics reports from companies like Foundation Medicine Caris Life Sciences and NeoGenomics Laboratories - and we were able to match hard-to-find patients in real time to precision trials. These parties never would have been able to find each other had we not deployed our tech and services into remote parts of the country and used AI to find those needles. We felt like we were saving lives every time we enrolled a patient, and I think we were. It was an incredible feeling, and an incredible outcome for our shareholders when we successfully sold the company in early 2022 to Paradigm.
The GenAI Unlock
Having said that, GenerativeAI is different. At Deep Lens and Predixion Software, we spent months building accurate ML or NLP models, which was great except that they only looked for one "thing" (like a mutation or test result type). To look at many, all or new "things" would require new models - a process which took months per "thing". They call this supervised machine learning. Generative AI is now capable of unlocking huge, heretofore-impossible-to-crack problems because it can look at all the data all the time (and new data as it comes in) and make sense of it in new ways.
At my new company Ryght, we have already seen massive lift in outcomes by applying this tech in specific copilots (apps) that enable knowledge workers at biotechs, big Pharma, CROs, RWD players and software vendors to speed up their day-to-day, but we have also developed copilots that can unlock millions of complex unstructured datasets in an automated pipeline that feeds the AI, dashboards, and knowledge graphs to solve big problems. Think of these as “data pipeline copilots”.
Copilots from out of the Permafrost
This unstructured data unlock is akin to the thawing out of a wooly mammoth from the permafrost of Siberia. Imagine interacting with a live mammoth that has thawed out after not seeing the light of day for 100,000 years.
We recently did just that, kinda ;-) We resuscitated some wooly mammoths by applying our generative AI data pipeline copilot to a library of trial document faxes and PDFs that had been stored in an EMR from an unnamed vendor. These documents contained images, hard to read charts and tables (some resembling printouts from a low-ink printer and some slanting sideways) - and with a specific technique learned by our code-free copilot builder we were able to not only extract the data from these documents with amazing accuracy (think of GenAI as OCR and ML on steroids - way more accurate and automatic) but the structured results were then automatically sent to our platform so that (in #ChatGPT style) users could query the data and output it to dashboards (think depositing it into repositories like @databricks and @snowflake and visualizing the data too).
Extrapolate that for a minute by imagining you are pointing to millions of PDF and fax documents, saying, "What are the most common gene variants in these data, where are the patients located, and what type of cancer do they have?" "Of this cohort, what are the most common test types and what tests were not run but should have been?". Then “Use this information to write inclusion/exclusion criteria for my protocol,” then “Output this data to a dashboard and make it available to my clinops team". The process outlined here is not only possible now, but think about it... without hundreds of people, years of ML development and only specific slivers of value being delivered, this has, until now, been impossible. Now you can query your faxes directly.
But You Need an Enterprise Architecture
At my ML platform company, Predixion Software (2009-2016 until its exit) I learned an important lesson: Customers don't care that your algorithm is 6% more accurate than the rest; They care about what they can do with this insight. They want to operationalize it, put it in the hands of knowledge workers who can move the needle for the company and get an ROI. At Ryght, we think the same way. To do Generative AI in the enterprise you need orchestration, security, compliance, scale, reporting, SLAs and Integration, and you need a platform that can support future copilots you haven't even thought of yet.
In case you missed it, read the last article in our series about the subject of orchestration here from our CTO, Johnny Crupi (who by the way was a distinguished engineer at THE Sun Microsystems and was CTO for their web services group (FKA SOA).
The future is upon us now, and the industry is ready to embrace it. Hold on for the ride.
To learn more about the intersection of Enterprise AI and Vertical SaaS in healthcare & life sciences at Ryght, or if you want to focus on specific aspects in more depth, contact us here.