Amazon
Software Engineer (current)
Items Security & Privacy: designed a denylist migration to AWS AppConfig.
ABOUT ME
Hi! I'm Grace, a software engineer and Cornell Computer and Information Science graduate. My background includes distributed systems at Amazon, healthcare software, and applied machine learning. I'm interested in what happens when things go wrong—and how the response changes what happens next.
Questions I'm exploring
How should AI respond when the stakes are high?
Exploring when mental-health AI should ask more, offer support, or defer to a clinician.
How can care continue beyond treatment?
Strengthening follow-up and ongoing support for people with substance use disorders.
What makes AI useful beyond a benchmark?
Understanding how models, interfaces, and clinical workflows shape whether AI helps in practice.
More about these questions
How should AI respond when the stakes are high?
When someone turns to AI in distress, what should it recognize, what should it ask, and when should it involve a human? How do we distinguish a response that sounds empathetic from one that offers appropriate support? I'm interested in how these decisions unfold across a conversation, and how to evaluate the harm a system might cause through both its actions and its omissions.
How can care continue beyond treatment?
How can people transition from residential treatment to everyday life without having to coordinate all their support themselves? I'm interested in how care coordination, relationships, and everyday environments can sustain support across that transition.
What makes AI useful beyond a benchmark?
Why might a model perform well in an experiment yet struggle in a clinical workflow? Is the limitation in the model, the information it receives, or how people interact with it? I'm interested in building experiments and tools that help distinguish these possibilities, and testing whether a more capable model is actually the improvement we need.
Where I show up
patient care
- Certified nursing assistant
- Medical assistant
- Phlebotomist
community support
- Crisis hotline volunteer
- Women's health advocacy
- Refugee advocacy
mentorship
- Cornell Techinterview & job-search coaching
- Cornell Bowers ASCENDAI & privacy project mentoring
A little more about me
My background includes work at Merck and the Englander Institute for Precision Medicine at Weill Cornell Medicine.
Caring for people one-on-one, at the bedside or on the phone, is where I find a lot of meaning. It keeps me grounded in the everyday realities of care and guides my interest in building tools that extend that support to more people.
Through Cornell Bowers ASCEND, I mentor students developing projects in applied AI and privacy, offering technical feedback and career guidance. I also help Cornell Tech students prepare for interviews and navigate the job search.
Outside of work, you'll usually find me reading, writing, making something in Adobe Creative Suite, or going for a run.
EXPERIENCE
Amazon
Software Engineer
Catalog Data Management: automated validation reports and led API and authentication migrations.
Honors
- Design Excellence Award, Biomedical Engineering
- National Merit Scholar
- AIME Qualifier (AMC 10)
PROJECTS
9 projects
what I'm working on now
safety evaluation for consumer mental-health AI
2026 · ongoing
Evaluating AI safety in mental-health crisis conversations with Stanford researchers, Grow Therapy, and ARISE, a research network founded by labs at Stanford Medicine and Harvard Medical School.
My contribution focuses on clinician voice-to-text integration and configuring simulated-patient behavior.
mental health prediction
2022
I used gradient-boosted models to predict mental-health scores from CrossCheck data for 61 participants. Excluding missing observations gave lower reported error than filling gaps with averages, but left only 17% of the original data for modeling.
python
minitorch
2021
I implemented autodiff, tensor operations, accelerated kernels, and convolutions in an educational ML framework, then used it for digit and sentiment classification.
- python
tiempo
2022
Helping undocumented New Yorkers check clinic wait times and book same-day appointments for non-emergency care.
care-home availability
2021
Making care-home availability easier to find. Our team prototyped SMS surveys and a staff dashboard using Twilio, cleaned availability data, and Python/Streamlit.
- python
neural activity modeling
2020
Our team compared linear, polynomial, and CNN-based predictions of neural activity across 940 neurons, testing how feature representations and group ablations changed the models' predictions.
- python
brain-computer interface
2020
A P300 speller for people with ALS: letters flash on screen, and a portable OpenBCI headset detects the brain's response when the intended letter lights up, so users can spell without moving.
Winner of the Design Excellence Award.
c++
perl
openvibe
depth estimation
2019
Estimating depth from thermal images with a fully convolutional neural network, supporting visualization of subterranean environments from a single camera.
- python
tensorflow
WRITING
Research, technical notes, and essays on health and technology.
4 of 6 writing items
publications
Affective Touch as Immediate and Passive Wearable Intervention
Zhao, Tao, Le, Maki, Adams, Lopes, Choudhury
Proc. ACM Interact. Mob. Wearable Ubiquitous Technol. 6, 4 (Dec. 2022)
An ACM IMWUT paper exploring affective touch as a just-in-time wearable intervention for in-the-moment anxiety.
Passive Targeting in Nanomedicine
Nanoparticles for Biomedical Applications, Elsevier (2019)
A chapter in Nanoparticles for Biomedical Applications on passive targeting, body interactions, and clinical potential in nanomedicine.
other writing
ART
A film edit, made from archived footage.
Watch on YouTube
Premiere Pro
Premiere ProLET'S CONNECT!
for collaborations, projects, or just a conversation, feel free to reach out through any of the platforms below or email me at dl2228@cornell.edu.