Camille Dunning

I am a first-generation senior undergraduate studying Data Science at the Halıcıoğlu Data Science Institute at UC San Diego. My research interests are long-range graph neural networks, deep learning-based structured signal recovery, and (inverse) reinforcement learning for robotics.

summary

During my time at UCSD, I was the recipient of two competitive scholarships offered by the Institute: the HDSI Undergraduate Research Scholarship and HDSI Industry-Sponsored Research Scholarship. I turned my focus to research after my Undergraduate Research Scholarship project, “Adaptive Quantitative Group Testing Using Deep Q-Learning and the Walsh-Hadamard Matrix”, mentored by Professor Tara Javidi, was picked as the best out of all projects of 20+ scholarship winners, and selected to be presented to the California Alliance for Data Science Education, based at UC Berkeley.

I am currently conducting research under the UCSD Astrophysics Department’s Cool Star Lab, in ML-based identification of ultracool spectral binaries, supervised by Professor Adam Burgasser. Around the same time I started this opportunity, I was working on training deep learning models on large-scale cyber-infrastructure for the San Diego Supercomputer Center. I am also working on a research project in inverse reinforcement learning at the Existential Robotics Lab, supervised by Professor Nikolay Atanasov and Tianyu Wang. Finally, I am mentored by Professor Yusu Wang in proposing a new multi-scale neural network architecture, based on clustering, GraphGPS, and sparse attention, that improves performance in tasks involving capturing long-range dependencies in graphs.

Despite my drive in research, my engagements in the industry have not slowed down. I built a proof of concept for a Snowflake data warehouse for BlackRock’s Aladdin Product Group, developed security capabilities for the Heroku platform with Salesforce, and created an end-to-end synthetic data generation and deep learning training pipeline for real-life scenes with NVIDIA. Currently, I am building ML pipelines and infrastructure for Tesla’s internal manufacturing applications. I have also previously worked on machine learning, deep learning, and cloud capabilities and research projects with the Scripps Research Translational Institute (supervised by Professor Giorgio Quer), RAPID AI, Tangible AI, and Blooma AI. I have also founded and served as a core backend, DevOps, and ML engineer for two startups, one of which participated in The Basement and StartR incubators. I still occasionally do freelance ML work for various venture-backed startups.

After graduating, I plan to work in the industry in a Data Science, Machine Learning, Solution Architecture, or MLOps role, before applying to MS and PhD Machine Learning or Computer Vision programs.

educational background

I am pursuing a B.S. in Data Science (3.83 major GPA) at UC San Diego, planning to graduate in June 2023. Some highlighted experiences and extracurriculars (fun and academic) include:

1st place winner of the advanced track, DataHacks 2021 (unanimous perfect score for my team), mentor for DataHacks 2022
Ranked second out of 40+ students in the Introduction to Data Mining course Kaggle competition.
HDSI Undergraduate Research Scholarship recipient (2021-2022). View the press release here.
HDSI Industry-Sponsored Research Scholarship recipient (2020). One of two selected for the award, out of hundreds of applicants. Received an offer for an internship with BD (rescinded due to COVID-19).
Co-Director (2021-2022) and Technical Lead (Fall 2022) of the Online Content Subcommittee of UCSD’s Data Science Student Society (DS3). View the committee’s Medium publications and podcasts here.
Authored the Sigmet Python library (2020) in a team of four in DS3, 1k+ downloads.
Authored Arpagen, a corpus and LSTM benchmark for phoneme-level text generation.
Complex systems educational internship with the Qualcomm Institute’s QI Learning Academy (Winter 2020).
2x participant (2019-2020, Fall 2021) in The Basement Incubator.

interested in my data science blog and podcast episodes?

The link here for your perusal.

news

Mar 15, 2023	Presented Graph-HSCN: Heterogenized Spectral Cluster Network for Long Range Representation Learning at the 2023 Halıcıoğlu Data Science Institute Senior Capstone Showcase! Check out the project on my Projects page to learn more.
Feb 1, 2023	Finally have an academic website. Construction is underway. 🚧

selected publications

Identifying Ultracool Binary Systems using Machine Learning Methods

Malina Desai, Juan Diego Draxl Giannoni, Camille Dunning, and 4 more authors

Research Notes of the AAS, Jan 2023

Abs

Ultracool spectral binaries are unresolved pairs of low-mass stars and brown dwarfs revealed by peculiarities in their combined light spectra. Methods to identify these systems have relied on spectral indices, which have known selection biases. We report on a pilot study examining the application of machine learning methods to identify ultracool spectral binaries. Using a sample of single and binary templates constructed from low-resolution, near-infrared spectra, we trained a random forest model to identify binaries composed of M7–L7 primaries and T1–T8 secondaries. We find that uniform data preparation and balancing of the training sample are critical to building an effective model. Our model achieves precisions of ≥95%, confirms known spectral binaries, and identifies new spectral features sensitive to multiplicity, illustrating the utility of machine learning methods to identify these rare systems.