Huma.AI Co-founders Lana Feng, CEO, and Greg Kostello, CTO, speak with Caroline Carruthers, Chief Executive at Carruthers and Jackson, in a video interview about their professional backgrounds and journey, the inception of Huma.AI, the importance of data accuracy in healthcare, the role of Generative AI as an enabler, need for human intelligence, challenges around AI, and how to tackle them.
Huma AI is a self-service business intelligence platform built for the healthcare industry. Caruthers and Jackson is an end-to-end data consulting solution that helps businesses harness the value of their data.
Feng begins by shedding light on her background. She was born into a family of physicians and ended up going to graduate school and obtaining a Ph.D. in biomedical research.
However, Feng states that she has always been an entrepreneur and prior to starting Huma.AI, she was hired to create a pharma business at Genoptics. She shares that the business successfully grew from zero to $45 million in just four years. As a result, Novartis acquired the company for $500 million.
As a subject matter expert, she discusses addressing the challenge of leveraging the right data to get better answers with Huma.AI and natural language processing. Feng notes that this empowers SMEs like herself to become data scientists and focus on building models and performing complex analyses.
Further, Kostello shares that he started working as a software engineer at Apple. Then, he became the CTO of a public company and along the way, created a large-scale SaaS platform used by ESPN, Cisco, Disney, and NASA.
Kostello has been working with large data since the late 90s in the consumer sector to assist those who require easy access to answers, without having a technical background.
Explaining the purpose behind the creation of Huma.AI, Feng says that her experience as a domain expert at Novartis prepared her to handle the increased volume of trials the organization has recently taken on. She has a background in precision medicine, specifically in targeted therapy for cancer.
According to Feng, data accuracy in this field is critical and it can impact the success of a trial. For instance, she says, if there is a 10% prevalence of a gene mutation, she may need to screen 1,000 lung cancer patients to find 100 with that mutation. However, if there is a mistake and the prevalence is only 1%, she would have to screen 10,000 patients to find those 100, making the trial likely to fail.
Although Feng recognizes the significance of data accuracy and knows the questions she wants to answer through the data, she lacks the technical knowledge to write SQL queries. This led her to wonder if they can leverage consumer capabilities in technology to empower domain experts like herself to become their own data scientists. The purpose of this platform is to accelerate the development of new drugs by leveraging data, says Feng.
When asked about negative press around generative AI, Kostello responds by saying that while there is some truth to it, it is also exaggerated. For him, it is important to always assume failure while working with generative AI.
However, Kostello states that this technology also provides answers that were previously unattainable. Because it is trained on human data, it understands human thinking, he adds.
Highlighting the complexity of massively unstructured healthcare data, Kostello maintains that having an understanding of this field is crucial for its application, without being completely reliant on the answers generated from that data.
Adding on, Feng mentions artificial intelligence and the fast adoption of ChatGPT with a billion users in 7 months. She states that as a life science practitioner, she approaches generative AI with caution and feels lucky enough to be working with OpenAI models.
Feng recounts that her guiding principle behind building a generative AI platform has been to ensure privacy and security to accurately analyze clients’ private data and be transparent to solve hallucination issues. She mentions that the company is on the right track based on the client’s response to the approach.
"COVID showed that medicines can be brought to market in 3-4 years, rather than a 10-year timeline."
Lana Feng | Co-founder and CEO, Huma.AI
Moving forward, she mentions how COVID showed that medicines can be brought to market in 3-4 years, rather than a 10-year timeline and a $3 billion price tag. However, the challenge lies in the massively complex healthcare data that is unstructured, and the industry relies on experts like herself to curate the data manually.
Citing an instance regarding the effectiveness of Huma.AI, Feng says that from 35 million publications in PubMed, the solution uses traditional NLP transformers to surface relevant publications and then uses Generative AI to give key takeaways and then bring in the citation. This allows experts to verify to see whether those generated key takeaways are valid and accurate.
According to Feng, this saves experts the task of reading the 100 papers that surface after putting in keywords. She affirms that as per the return-on-investment studies, clients confirmed deriving value in minutes instead of months.
"Prioritizing human intelligence every step of the way is crucial for us."
Greg Kostello | Co-founder and CTO, Huma.AI
In addition, Kostello states that the other crucial thing is prioritizing human intelligence every step of the way. He shares that the final output is reviewed and used by human beings, physicians, and scientists to analyze the data.
In accordance, Feng says that AI will not replace humans, but humans who use AI will replace those who do not. She looks at AI as a tool to empower experts and poses the question of how one can leverage AI to solve real-world problems. Speaking of challenges, she stresses trust issues and the fear of AI taking away jobs, which goes back to the education piece.
Furthermore, she states that the aspect of build versus buy has become the C-Suite discussion. Feng mentions discussing how every company requires a Generative AI position, the use cases to leverage Generative AI, and bringing value within the next six months.
In conclusion, Kostello states that timing is everything. He recalls convincing people initially that the NLP interface was the right thing until Chat GPT took off. He shares that while building the company, they built in verification, validation, and transparency from day one, which helped alleviate many concerns.
CDO Magazine appreciates Lana Feng and Greg Kostello for sharing their insights with our global community.