David Aponte is a machine learning engineer, working in machine learning infrastructure and ML-Ops. With a background in molecular biology, education, and data science, he enjoys building solutions to support multidisciplinary domains.
He also helps lead and organize the ML-Ops Community, a global community of ML-Ops practitioners dedicated to sharing and promoting best practices for machine learning in production.
Summary
In this episode, David starts off by giving us a peek into his day-to-day as a machine learning engineer. He gives us an insider view of the drug discovery domain and how AI/ML is helping with real-life examples of the developments during the pandemic.
David is passionate about the focus on data and emphasizes it as the fuel that drives the AI/ML engine – the more the better, the higher quality fuel there is, the better access that one gets will all result in the best outcomes.
Audience Questions
[00:36:00] Can AI offer decision support in balancing treatment choices between short-term-duration with medicines and intervention etc. vs. long-term-duration with primary care and lifestyle changes?
Alberto Rambaudi — Health System Integration, Italy
[00:38:55] What technology choices did you make when you started your project – what would change if you were to start today?
Ramasamy Apathotharanan — Senior Manager – R&D at Aruba, a Hewlett Packard Enterprise company, Bangalore
Topical Questions
[00:07:36] In your bio you mentioned that you have a background in molecular biology as well..how do you think that background is helping you?
[00:09:00] Give us a little bit of a peek into how machine learning is helping drug discovery? What is the potential in this space?
[00:12:55] Do you think David that, because of the COVID-19 pandemic – the amount of money that’s coming into, maybe drug discovery companies has increased many fold or was this something that was any way happening?
[00:14:35] …there was also a lot of flak given to the AI community that they were not, with all the talk about AI and hype about AI, that it really was not able to let’s say engineer a treatment or discover a treatment for COVID-19…how do you respond to that?
[00:17:55] If you had a million dollars to invest, which are the areas within healthcare, like drug discovery, where you would invest your money in and why? Especially AI in Healthcare, which areas do you think are really hot?
[00:27:54] ML-Ops in a healthcare setting or deployment, what’s different about that?
Quotable Quotes
[00:08:02] You can’t avoid the domain, you can’t avoid the subject that you’re working on, as much as we’d like as engineers to kind of just abstract away all of that, and just focus on like a technical problem. A lot of the time, you have to work with other people that are experts. So, it’s understanding what are their concerns, what are their cares.
[00:10:00] On average, it takes like 10 to 14 years and billions of dollars to get a drug into production..what we’re trying to do is speed up that process, we’re trying to make that more efficient, make it more scalable and in many ways, automating things that can be easily automated.
[00:16:50] Machine learning is not only becoming more ubiquitous, but it’s also becoming more mature. It’s becoming something that you’re going to see embedded into real life processes and real time even too.
[00:19:30]…in my experience, most of the challenges and the tech healthcare space are still around data. How can we get good data? How can you get reliable data? And how can we get that reliable data to the applications that need it, whether it’s in batch, whether it’s in real time, whether it’s for research or whether it’s production systems, getting that data, having being able to discover new data sources is really the fuel of machine learning.
[00:31:23] Biology is really hard and when you’re training a predictive mechanism — you train to predict whether or not this gene causes that disease, you’re trying to generate compounds that can hit that target; they are high stakes..the consequences of getting that wrong are expensive and potentially harmful…
Notable Mentions
[00:10:40] Use of Baricitinib
[00:22:02] ML-Ops Coffee Sessions Podcast
Pointers to past You+AI Podcast episodes
[00:34:20] S1-E15 Demystifying Immunity, with Revati Masilamani
Connect
LinkedIn: David Aponte
The You+AI Vodcast
A companion video segment full of fun and candid moments.
Check it out here!