Julia Kim

PhD student @ MIT ORC

Hi, I’m Julia! I’m a second-year PhD student at the MIT Operations Research Center, advised by Professor John Horton. My research interests include multimodal AI, automated research, and AI for financial decision-making. I’m currently working with John Horton and The Carlyle Group on AI agents that reconstruct and evaluate investment decisions using the information available at the time.

Before MIT, I studied mathematics at Cambridge and physics, mathematics, and statistics at the University of Toronto. I’ve also worked in The Trade Desk’s AI Lab. You can find more information in my CV.

Julia Kim

Research Highlights

I develop and evaluate AI systems for learning and decision-making. Here are some of my recent projects.

see what I’m working on
ⓘAn example of visual explanation: activations propagate through a neural network. Animation by Grant Sanderson / 3Blue1Brown; illustrative inspiration, not output from our platform. Original lesson.
Manuscript in preparation, 2026

Personalized Learning with Generative AI: A Framework for Assessment-Driven Remediation Videos

Julia Kim, Dimitris Bertsimas

We combine reusable remediation videos with personalised visual annotations. Missed questions select the lesson; incorrect answers and optional tutoring dialogue select clarifications shown during playback, without changing the narration or video.

demoproject
ⓘWes Cockx / Google DeepMind / Pexels. Artist’s interpretation of language models from Visualising AI; not PortIA output.
MIT 15.090 course project, 2026

PortIA: Language Models for Portfolio Optimization

Julia Kim, Lindsay Carlin, Jacob Dentes, Salvador Dzimah

PortIA translates natural-language investment preferences into portfolio constraints, then uses mathematical optimization to construct portfolios. The project combines language models, market data, and sparse portfolio selection.

slidescode
ⓘNASA, ESA, and F. Summers, G. Bacon, Z. Levay, J. DePasquale, L. Hustak, L. Frattare, M. Robberto, M. Gennaro (STScI), R. Hurt (Caltech/IPAC), M. Kornmesser (ESA). Acknowledgement: A. Fujii, R. Gendler. Orion Nebula fly-through at 1.25× speed; illustrative context, not StarCLIP output.
Cambridge Part III essay, 2025

Applying Contrastive Learning to Stellar Spectra

Julia Kim, Nathan Sandford

StarCLIP brings spectra from different instruments into a shared representation, connecting contrastive learning with the physical properties of stars.

papercode
ⓘNASA, ESA, G. Bacon, L. Frattare, Z. Levay, F. Summers (Viz3D Team, STScI), and J. Anderson (STScI). Westerlund 2 fly-through, played at 1.25× speed. Illustrative visualization of a different star cluster, not observations of Omega Centauri or red-giant pulsations.
JAAVSO, 2022

Pulsating Red Giants in a Globular Cluster: ω Centauri

Julia Kim, John R. Percy

We analyse the light curves of 16 pulsating red giants in ω Centauri, studying their pulsation periods, long secondary periods, and changing amplitudes.

paperabstract
a little catch-up ☕...

News + Updates

I’m working with Professor John Horton and The Carlyle Group on AI for investment decision-making.

milestone

I started my PhD at the MIT Operations Research Center!

milestone

Misc

  • I enjoy learning languages: I’m fluent in French and learning Turkish.

  • I’m interested in aesthetics, from the gym to skincare and beauty.

  • I also enjoy digital design from time to time.