Suggestions
Carmen Lai
Research @ Stealth AI Startup
Carmen Lai, Research Data Scientist at Stealth AI Startup
Carmen Lai is a full-stack data scientist with over 10 years of industry experience and a track record of success.2 She is currently working as a Research Data Scientist at a stealth AI startup.125
Lai holds a degree from Duke University2 and has a background in neuroscience with a passion for biotechnology.3 Her work experience includes applied machine learning, clinical genetics, and helping companies like Color implement precision healthcare programs at scale.34
In addition to her research work at the stealth AI startup, Lai is also involved in the AI and Computational Biology community.1 She is known for her expertise in areas such as automatic differentiation and computer algebra.1
Highlights
Happy Friday! I wrote a blog post with Edge Analytics Inc on interpretability tools for understanding ML models (SHAP values, decision trees, linear models), with code in a Colab notebook that you can run to reproduce the plots on an example dataset.
My interest in ML interpretability started from the application side (working with medical geneticists!), so this was a helpful (and more challenging than expected) exercise to formalize my learnings. I also discovered many helpful resources online which are linked in the post.
I hope this will be useful to other data scientists developing ML systems. Would love to hear any thoughts or additional resources you have to share!
My first data science press release! I wrote an article months ago about the importance of widespread COVID-19 testing. It's amazing to now get to work with data from large-scale testing at Color that validates this point, particularly for a more diverse population outside of the hospital setting.
Key takeaways based on a cohort of 30,000+ people:
-
A variety of symptoms were reported among positives and only a small proportion (12%) reported a high fever. ⇒ Fever screening and temperature checks may miss the large majority of positive cases.
-
A large proportion of positive cases (30%) reported no symptoms at the time of testing. ⇒ Symptoms screening may not be sufficient in containing an outbreak given a high presymptomatic/asymptomatic rate among positives.
-
The majority of people who tested positive are young (68% ages 18–40), and the data confirmed disparities in positive results based on ethnic background.
To safely reopen and effectively prevent an outbreak, we need more comprehensive testing (exposure-based testing, or proactive testing for everyone) regardless of symptoms.
Read the press release here: https://bit.ly/2C6Ty72
#covid19 #datascience
