Research

When One Discipline Isn't Interdisciplinary Enough

Research.
Astrobionano...
something

When I was just 7 years old, I recognized I wanted to spend my life in research & development because I knew that this field would always allow me to pursue the cutting edge of new technologies that hadn’t had the chance to collect dust. I wanted to not just build the next year’s work, but the next decade’s work. And throughout my journey, my thirst for research has only expanded.

Publications

ORCID iD iconhttps://orcid.org/0000-0002-8318-2596

2024

Characterizing Training Performance and Energy for Foundation Models and Image Classifiers on Multi-Instance GPUs


PDF | DOI

GPUs are becoming a scarce resource in high demand, as many teams build and train increasingly advanced artificial intelligence workloads. As GPUs become more performant, they consume more energy, with NVIDIA's latest A100 and H100 graphics cards consuming upwards of 700W of power. This paper characterizes how best to scale down a large modern GPU to suite workloads that cannot fully exploit an entire GPU. The paper measures six workloads from 14 million parameter image classifiers to 1.5 billion parameter large language models, finding that partitioned GPUs with a mix of small, medium, and large partitions can deliver up to 33% less energy demand and 9% higher training throughput from a single GPU. We found high potential in fine-tuning existing models, with 55% faster training at 42% less energy. Our results suggest that multiplexing small workloads onto spatially partitioned GPUs can improve the efficiency of a single GPU while giving clients access to smaller slices of the latest GPUs that better suits their job's demands.

2021

Improving Healthspan through Patient-Derived Artificial Organs from Induced Pluripotent Stem Cells and Two-Photon Polymerization


PDF | UPenn

Senescence, from the shortening of telomeres, accumulation of mutations, epigenetic hypomethylation, and other causes, begins an eventual cycle of decline in every patient. Stem cells disappear as a function of age, which in turn impairs cellular replication. Moreover, when fully differentiated cells are induced back into induced pluripotent stem cells (iPSCs), they not only revert to a state of pre-differentiation, but also to a younger cellular age. Their aging clocks turn back: their telomeres become longer and DNA methylation reverts back to an earlier age. It should therefore be possible to use iPSCs to replace the missing stem cells from aged organs and tissues and to replace existing, older tissue with younger cells. In order to accomplish this, this paper will also explore new nanotechnological mechanisms in generating patient-specific scaffolds, including two-photon polymerization, a fabrication process that uses a specifically focused, near-infrared laser to build three-dimensional scaffolds. This paper finally proposes a mechanism to apply iPSCs as preventative medicine, to replace aging organs before they fail and unwind the aging clock to lengthen healthspan.

2020

An Automated Approach to Modeling Jupiter’s Synchrotron Radiation from Radio Telescope Observations


PDF | DOI

After taking an astronomy course in the fall of freshman year, I then worked in a group of two other students to publish our paper on synchrotron radiation with NASA's Jet Propulsion Laboratory in her Astronomical Research Seminar. In our group, our instructor learned along with us as we analyzed models of synchrotron radiation—light emitted by charged particles moving in a curved path—for NASA’s Juno satellite instead of performing the typical analysis of a double star or exoplanet transit. However, more important than the actual science, I learned through that course, my instructor, and my teammates how to conduct scientific research going forward.

After presenting at NASA's Citizen Science Conference (shown above) in June 2019, we were published in the peer-reviewed journal Astronomy Theory Observations and Methods in July 2020. Our presentation was also cited by NASA's Jet Propulsion Labortory in Velusamy et al's "Goldstone Apple Valley Radio Telescope Monitoring Flux Density of Jupiter's Synchrotron Radiation during the Juno Mission."

2020

Comparing the Ages of NGC 1513 and NGC 2420


PDF | p. 137-140

The next year, my astronomy instructor asked me to be a teaching assistant for that seminar. In the process, I led two other groups (which developed this publication and "Observation and Investigation of NGC 1662" below), both focused on analyzing star clusters. Throughout this process, I would mentor these teams once a week—directing each team by helping them find these questions they did not know they needed to ask. In particular, one group considered using a Python script to learn more about the stellar composition. I then discussed with them the goals of the project, and we realized that the script was tangential to our focus. After pivoting, I then developed other tools in Python and helped generate their figures. Most importantly, I brought my experience from the previous year and led these teams to publication by guiding them to develop a purpose, focus on the relevant information, and discover what questions need answering.

2020

Observation and Investigation of NGC 1662


PDF | p. 93-98

As noted in the above publication "Comparing the Ages of NGC 1513 and NGC 2420," this was another publication where I used my previous research experience writing and publishing "An Automated Approach..." to mentor a research team discovering double stars.

Also Explore

Computer science was an early passion, as such, I have developed many technical projects exploring various fields. As I learned more, I have become fascinated with how I can use technology to find medical solutions, whether detecting diseases early or creating preventative cures. To this point, the various technical projects that have developed my computer science background include,

Technology