What Is the Polo Club of Data Science at Georgia Tech?

Updated

The Polo Club of Data Science is a research group at Georgia Tech, based in Atlanta, Georgia, that builds scalable, interactive, and interpretable tools to help people understand and interact with billion-scale data and machine learning models. It is relevant to you if you want to follow or understand research at the intersection of data visualization, graph mining, human-centered AI, and security — or if you are looking for a lab whose output includes both papers and publicly usable visualization tools.

What the group works on

The club describes four current research thrusts:

  • Human-centered AI — interpretable, fair, and safe AI; adversarial machine learning; interactive intelligent visualization; scalable discovery of deep learning models; and AI education for students and educators.
  • Large graph visualization and mining — making sense of very large networks and dimensionality reduction through network science.
  • Cybersecurity and fraud — including work connected to identity security and adversarial machine learning.
  • Social good — applications in areas such as health and energy.

The unifying goal across these areas is tooling: systems that scale to billions of data points, respond interactively, and explain what a model is doing rather than acting as a black box.

Who is involved

The group is led by Polo (Prof.) and includes current members and recent alumni across degree levels — CS PhD, ML PhD, CSE PhD, CS MS, and CS undergraduates. Alumni have gone on to roles at organizations including OpenAI, Stripe, Apple, Meta, Figma, JPMorgan Chase, and academic positions such as Cal Poly, as well as PhD programs at Stanford, MIT, and CMU. This mix of levels means the club functions as a lab with both long-term doctoral research and shorter-term undergraduate and master's contributions.

Notable recent work

The group's news page lists accepted papers and awards, which is the most concrete way to see what it currently produces:

Project Topic Venue / recognition
UNIPO Visualizing how GRPO, DAPO, and other RL algorithms fine-tune LLMs, token by token IEEE VIS'26 Best Short Paper, Honorable Mention
Flow Reasoning Models Turning flows into efficient recurrent reasoners NeurIPS'26 DiffuLM Workshop
Time Conditioning Optional for Flow Matching How data geometry makes time conditioning optional NeurIPS'26
RECAP RL post-training that teaches reasoning models to override flawed thinking EMNLP'26 Main
ComplicitSplat 3D-related work BMVC Main Conference
UMAP kNN Graph Understanding dimensionality reduction through network science TopoInVis@VIS'26

The group also engages outside academia — for example, an invited participant at an Atlantic Council roundtable on identity security and national resilience, and a keynote on realistic simulation as a lens into computer vision vulnerabilities.

How to use this

  • If you want to follow the research: the group's site is the primary source for papers, tools, and news, and several projects ship as interactive visualizations rather than papers alone.
  • If you are evaluating it as a lab: the alumni placement record and the range of venues (VIS, NeurIPS, EMNLP, BMVC) give a sense of where graduates land and what kinds of work the group publishes.
  • If you only need one takeaway: the Polo Club is a Georgia Tech data science lab focused on making large-scale data and ML models understandable, with security and social-good applications as recurring themes.