Newport Beach, CA

Gia Ancone

Pattern-mapping our world at the intersection of AI and Physics

Education: Stanford University
B.S. Computer Science (AI), Math Minor - 2027
M.S. Computer Science (AI) - 2028
Where: Stanford NeuroAI Lab
SLAC Julia Group
Kavli Institute for Particle Astrophysics & Cosmology
CURRENT FOCUSES: Computer Vision for Robotics World Modeling
Physics Foundation Models
Galaxy Cluster Relaxation

Research and Experience

012024–Present

World Modeling for Robotics & Object Understanding

Stanford NeuroAI Lab

ADVISED BY DANIEL YAMINS, KLEMEN KOTAR, ATLAS KAZEMIAN & RAHUL VENKATESH

We're currently training PSI, the NeuroAI Lab's autoregressive world model, to advance generalized robotic understanding and prediction. Previously as a 2025 Stanford CURIS fellow, I probed object representations in visual world models, building motion-grounded benchmarks for segmentation, material understanding, and affordance discovery.

CVPR 2026 PAPER SpelkeNet CURIS POSTER

02Summer 2026

Hardware-Aware Foundation Model for Edge Waveform Analysis

SLAC National Accelerator Laboratory · Julia Group

ADVISED BY JULIA GONSKI

Incoming summer researcher with the Julia Group at SLAC National Accelerator Laboratory. Awarded the Stanford VPUE Major Grant in support of this project.

032024-Present

Galaxy Cluster Relaxation

X-Ray Observational Cosmology Lab & KIPAC

ADVISED BY ADAM MANTZ & ARTEM POLISZCZUK

Improved the purity of predicted relaxed galaxy cluster samples using accessible photometric data and X-ray centering observations. Developed an algorithm to identify BCGs missed by red sequence-based cluster member catalogs, improving the reliability of optical relaxation metrics.

04Fall 2025

Probabilistic & Unsupervised Machine Learning

Oxford University · Magdalen College Visiting Stident

ADVISED BY SCOTT LE ROUX

An intensive Oxford tutorial on probabilistic and unsupervised learning and deep generative models: diffusion models, VAEs, and normalizing flows, from mathematical theory through implementation.

CONSISTENCY MODEL PAPER

05Winter 2026

CausalT5k: A Diagnostic Benchmark for Causal Reasoning in LLMs

Publication from CS 372

ADVISED BY EDWARD CHANG

A diagnostic benchmark of 5,000+ cases across 10 domains testing whether LLMs detect rung collapse, resist sycophantic drift under adversarial pressure, and generate "Wise Refusals" when evidence is underdetermined.

KDD 2026 PAPER

062023–25

Fountain Hopper Rocket Controls

Stanford Student Space Initiative

Controls Co-Lead · Software Engineer

Co-led the controls team for SSI's first self-landing rocket, and wrote the real-time, in-flight moment-of-inertia code that underpins active stabilization during flight.

Projects

Selected

01Winter 2026

Adaptive Inference Routing for Autoregressive-Diffusion Generative Models

CS 229, MACHINE LEARNING · FINAL PROJECT

A router that dynamically chooses between AR and diffusion paths in generative models, spending compute on quality only where it actually matters.

Project Poster

02Spring 2026

Learning Structured Trust Policies from Uncertainty, Advisor Signals, and Agreement

CS 224R, DEEP LEARNING · FINAL PROJECT

Learning when and whom an agent should trust via reinforcement-learned trust policies.

Project Poster

Let's build together.

Email gancone@stanford.edu LinkedIn /in/gia-ancone Google Scholar research profile Résumé download PDF