Undergraduate Researcher

CleAR Lab · UT Austin · June 2024 – May 2026
Out-of-distribution detection · Multi-agent planning

At CleAR Lab I work on applied ML for autonomous safety — from video out-of-distribution detection with world models to game-theoretic multi-agent planning. I am a second author on a paper accepted to AAMAS 2026 and TMLR (arXiv:2505.00213).

  • Extended SCOPED from static images to multiclass video OOD detection using temporal representations from NVIDIA Cosmos-Predict2.5, achieving 99% AUROC on the Honda Driving Dataset
  • Built a JAX research framework integrating Motion Transformer predictions with a mixed complementarity problem solver for safety-aware multi-agent planning on the Waymo Open Motion Dataset; designed a decoder to select relevant agents from multimodal trajectories
  • Developed a game-theoretic multi-agent path-planning algorithm in Julia using a Player Selection Network game, reducing runtime 10×

Tools: JAX, Julia, PyTorch, Cosmos-Predict2.5, WOMD, Game Theory

CleAR Lab — autonomous planning research

Cross Mission Effectiveness Intern

The Aerospace Corporation — El Segundo · May 2026 – August 2026
LLM evaluation · RAG · Classical ML

  • Designed an LLM-as-a-judge experimental benchmark with 30 expert-authored cases and reference-guided grading to evaluate 4 open-source models across 5 configurations
  • Evaluated Claude Sonnet 4.5 for integration with government modeling-and-simulation software and built a RAG-assisted AI workflow that generated and validated an AFSIM radar model, improving target detection 2.6×
  • Engineered a feature pipeline for 5M+ noisy AIS observations, filtering invalid vessel trajectories, and trained a scikit-learn logistic-regression classifier for military vessel identification

Tools: LLMs, RAG, scikit-learn, Python, AFSIM

Machine Learning and Entertainment Engineering Co-op

ATA Engineering — San Diego · August 2024 – December 2024
Diffusion models · Photogrammetry

  • Implemented a conditional diffusion model in PyTorch for hypersonic heat-flux prediction, running 100× faster than CFD; presented findings to 40+ ML researchers
  • Led development of a structure-from-motion photogrammetry workflow for fluid analysis, reconstructing wind-tunnel finite-element meshes from only 2 sample images
  • Developed Python tools for automatic VTK mesh segmentation and research-data management using Git

Tools: PyTorch, Diffusion Models, Photogrammetry, Python, VTK

Diffusion model architecture for hypersonic heat flux prediction Model architecture used. Real photos proprietary.

Aerospace Analysis Engineering Co-op

ATA Engineering — San Diego · May 2025 – August 2025
Learning-based control · Structural & thermal analysis

  • Formulated a real-time nonlinear model predictive control (NMPC) algorithm with ML-based state estimation in PyTorch for robust liquid rocket engine valve control
  • Led solar-loads analysis in STAR-CCM+, including geometry idealization, meshing, HPC parallelization, customer risk mitigation, and technical reporting
  • Built a MATLAB load signal generator for NASTRAN modal-transient structural analysis of VTOL aircraft motors

Tools: PyTorch, NMPC, STAR-CCM+, NASTRAN, MATLAB, HPC

ATA Engineering — aerospace analysis work

Lead Software Engineer

Texas Aerial Robotics · UT Austin · September 2023 – May 2026
Robotics · Autonomy · Control

  • Led 9 engineers in building and validating a C++/Python ROS 2 autonomy stack for object detection, navigation, and aerial retrieval in PX4/Gazebo; achieved 100% success in a university demonstration
  • Built a real-time receding-horizon control algorithm for centralized trajectory optimization of 4 drones, enforcing dynamics and collision constraints in simulation
  • Developed MATLAB UAV simulation software for the 2024 Raytheon drone competition, implementing A* pathfinding on an NCAA football field

Tools: ROS 2, Gazebo, C++, Python, OpenCV, PX4, MATLAB

TAR Drone — Flight Demo

Chief Engineer — Senior Capstone

FlareX Wildfire Competition · UT Austin · August 2025 – Present
Vehicle design & dynamics

As Chief Engineer, I lead technical development of a VTOL-capable autonomous wildfire-response aircraft. My current focus is a custom transition simulation for the VTOL-to-forward-flight window, coupling vehicle dynamics with propulsion power budgets and stochastic wind perturbation models to optimize propeller throttle curves.

  • Lead engineer across structures, propulsion, and controls while meeting system requirements
  • Developing a VTOL-to-forward-flight transition simulation with vehicle dynamics, power requirements, and stochastic wind models

Tools: MATLAB, Innoslate, SolidWorks, Flight Dynamics, Systems Engineering

MPC servo payload bay assembly — CAD render MPC Servo Payload Bay — Senior Design

Nosecone Responsible Engineer

Texas Rocket Engineering Lab (TREL)
Structures & separation systems

I owned structural integrity and fabrication of the rocket nosecone: analytical stress calculations on the bulkhead interface, validation via ANSYS FEA, and leading a five-person team on a mechanical separation system rated for 16,000+ lbf·in of flight load.

  • Executed bulkhead stress calculations and correlated with ANSYS FEA for the nosecone–fuselage joint
  • Directed design of an integrated separation system (pyrobolt + mechanism) for 16,000+ lbf·in flight load
  • Post-processed and laid up a full-scale composite nosecone for structural and stage testing
  • Authored CDR for the pyrobolt and separation systems to secure design approval

Tools: ANSYS FEA, Stress Analysis, Composites, CDR Authoring, GD&T

TREL — nosecone structural work