cv

Exchange Scholar

National University of Singapore

• Awarded a merit-based placement through a highly competitive, university-wide application process.
• Graduate AI Coursework: Neural Networks and Deep Learning II
• Graduate Neuro Coursework: Frontiers in Neurotechnology, Behavioral & Cognitive Neuroscience.
• AI Coursework: AI Planning and Decision Making, Mind and Machine
• Neuro/Bio Engineering Coursework: Bioinformatics, Organoid Engineering, Mind and Machine

Bachelor of Science in Computer Science

University of Wisconsin-Madison

• GPA: 3.81/4.0
• Honors: Letters & Science Dean's List, Honors in Computer Science Program, Letters & Science Honors Program
• Graduate AI Coursework: Foundation Models, Learning-Based Image Synthesis, Advanced Robotics, Advanced Reinforcement Learning, Theory of Multi-agent Machine Learning.
• AI Coursework: Directed Study on LLMs with Dr. Sala, Game AI, Neurobiology, Artificial Neural Networks, Artificial Intelligence, Game Theory and Learning.
• Neuro Coursework: Molecular and Cellular Mechanisms of Memory, Neurobiology.

Foundation Models Researcher

Dr. Sala's Sprocket Lab

• Researching and implementing novel methods for editing representations within large language models including coarse-to-fine pipelines.
• Worked with big data, GPU clusters, and high throughput systems.
• Publications from this research: [7], [8]

Founding Director & Research Lead

Wisconsin Neuromorphic Computing and NeuroAI Lab

• Established and direct a university-sanctioned entity to explore the intersection of neuroscience and artificial intelligence.
• Secured formal funding, dedicated space, and administrative support from UW-Madison.
• Delivered lectures on advanced topics intersecting neuroscience and AI to audiences of over 100.
• Mentoring researchers from initial project proposals to research papers.
• Publications from this research: [2], [4], [6]

Reinforcement Learning Researcher

Dr. Hanna's Badger RL Lab

• Designing and implementing reinforcement learning policies to train physical robots for autonomous soccer, focusing on multi-agent coordination.
• Key Accomplishments: Top ranking in RoboCup International Robotics Competition (3rd place in the Standard Platform League 2025; 1st place in the Challenge Shield League 2024).
• Publications from this research: [1] (in collaboration with Dr. Berland's Lab), [5]

Reinforcement Learning Researcher

Dr. Berland's Complex Play Lab

• Architecting communication strategies between agents in multi-agent reinforcement learning (MARL) environments.
• Researched and developed a novel adversarial co-evolution framework to automatically generate curriculum for MARL.
• Publications from this research: [1] (in collaboration with Dr. Hanna's Lab), [3]

Software Engineer (Research Engineer)

Stealth Mode Startup

• Spearheaded the complete research and development lifecycle for a novel artificial intelligence system, taking the project from an ambiguous high-level goal to a fully deployed, production-ready system.
• Devised, prototyped, and implemented a custom, hardware-aware algorithm that significantly outperformed SOTA approaches by over 100x in accuracy on the system's hardware.

Quantum Software Intern (Research Engineer)

HRL Hughes Research Laboratories

• Architected a multi-pass compiler in Common Lisp to generate optimized binary directly for a custom quantum control processor (the QICK tProcessor ISA), creating a low-level pathway for direct FPGA execution to maximize performance and control flexibility.
• Designed and implemented a custom assembly language to bridge high-level experimental logic with the hardware instruction set, enabling advanced compiler optimizations, automatic resource allocation, and precise picosecond-level timing calculations.
• Awarded a return offer after each period with HRL in recognition of significant technical contributions and research impact.

Quantum Software Intern (Research Engineer)

HRL Hughes Research Laboratories

• Engineered a production-ready implementation of an exact pattern matching algorithm within the Quilc quantum compiler, translating a novel theoretical method into a high-impact optimization tool.
• Reduced quantum circuit depth by up to 37%, shortening execution time on quantum hardware, directly enhancing algorithmic fidelity by mitigating qubit decoherence.
• Project Link: https://github.com/quil-lang/quilc/tree/master/src/match
• Awarded a return offer after each period with HRL in recognition of significant technical contributions and research impact.

[1] Communicating Plans, Not Percepts: Scalable Multi-Agent Coordination with Embodied World Models

Brennen A. Hill , Mant Koh En Wei , Thangavel Jishnuanandh

In proceedings of NeurIPS 2025 Workshop on Scaling Environments for Agents; in NeurIPS 2025 Workshop on Embodied World Models for Decision Making; and in NeurIPS 2025 Workshop on Optimization for Machine Learning

Developed a novel intention communication framework in MARL that achieved >96% success in a complex coordination task using a learned world model for latent trajectory planning and a self-attention mechanism to encode and share agent intentions, significantly outperforming emergent protocols.

[2] The Geometry of Cortical Computation: Manifold Disentanglement and Predictive Dynamics in VCNet

Brennen A. Hill , Zhang Xinyu , Timothy Putra Prasetio

In proceedings of NeurIPS 2025 Workshop on Symmetry and Geometry in Neural Representations and in NeurIPS 2025 Workshop on Interpreting Cognition in Deep Learning Models

Designed VCNet, a novel architecture emulating the primate visual cortex, achieving state-of-the-art accuracy on two vision benchmarks with over 10x greater parameter efficiency than standard models.

[3] Co-Evolving Complexity: An Adversarial Framework for Automatic MARL Curricula

Brennen A. Hill

In proceedings of NeurIPS 2025 Workshop on Scaling Environments for Agents

Developed a novel adversarial co-evolution framework to automatically generate a curriculum for multi-agent reinforcement learning that induces complex emergent strategies, increasing agent task performance by over 300% compared to baseline.

[4] The Physical Basis of Prediction: World Model Formation in Neural Organoids via an LLM-Generated Curriculum

Brennen A. Hill

In proceedings of NeurIPS 2025 Workshop on Scaling Environments for Agents and in NeurIPS 2025 Workshop on Embodied World Models for Decision Making

[5] Hierarchical Task Environments as the Next Frontier for Embodied World Models in Robot Soccer

Brennen A. Hill

In NeurIPS 2025 Workshop on Embodied World Models for Decision Making

[6] Structural Plasticity as Active Inference: A Biologically-Inspired Architecture for Homeostatic Control

Brennen A. Hill

In review

Developed a novel architecture that integrates synaptic and structural plasticity, demonstrating that computational agents can solve tasks by physically migrating their processing units on a grid to minimize local prediction error, driven solely by an intrinsic, active inference-based objective.

[7] HEFT: A Coarse-to-Fine Hierarchy for Enhancing the Efficiency and Accuracy of Language Model Reasoning

Brennen A. Hill

In review

[8] Breaking to Build: A Threat Model of Prompt-Based Attacks for Securing LLMs

Brennen A. Hill , Surendra Parla , Venkata Abhijeeth Balabhadruni , Atharv Prajod Padmalayam , Sujay Chandra Shekara Sharma

In review

Representation Fine-Tuning for Vision-Language Models

• Investigated Representation Fine-Tuning, a parameter-efficient fine-tuning method, on a vision-language model (nanoLLaVA) for a spatial reasoning task.
• Co-authored a research paper demonstrating that ReFT achieved accuracy comparable to LoRA (65.7% vs. 66.0%) while using nearly 10x fewer trainable parameters (0.019% of the model).

Reinforcement Learning for Quadruped Roller Skating

• Co-authored a research paper on training a Unitree Go1 quadruped robot with passive wheels to skate using reinforcement learning in the Isaac Gym simulator.
• Developed an RL policy that resulted in emergent complex behaviors, including automatic gait switching from a stable diagonal gait at low speeds to a dynamic galloping gait at high speeds (3 m/s).

Presenter

National University of Singapore School of Computing Showcase

• Selected to present a self-developed 3D videogame to an audience of over 100 students and faculty.
• The game featured AI-driven monster agents that used complex pathfinding algorithms to navigate dynamic 3D environments and make strategic targeting decisions; complex player abilities including harvesting resoucres, crafting equipment, and placing structures; and representing a unique mix of PvE, survival, resource management, and tower defense.

Winner (awarded $2,500)

Hack Midwest 2024

• Won (of over 300 developers), awarded $2,500, and noted for Best Enterprise-Scale Buisnesss Solution.
• Designed and built Badger Vision, an AI-powered assistive tool to help individuals with prosopagnosia (face blindness) by providing real-time audio cues for face identification and emotion recognition.
• Implemented a computer vision pipeline using deep learning (convolutional neural networks) to identify individuals and classify their emotional expressions from a live low-level video stream.

Honorable Mention

NASA International Space Apps Challenge 2024

• Received a global Honorable Mention, only awarded to 19 of 93,520 (0.02%) global participants.
• Processed large-scale NASA/ESA astronomical data and implemented the 3D visualization, rendering, and user interface.
• Won the Chicago hackathon and advanced to global finals; additionally awarded Best Presentation.

3rd Place (Standard Platform League 2025), 1st Place (Shield 2024)

RoboCup International Robotics Competition

• Achieved top placements in a competition that serves as an international scientific benchmark for multi-agent AI in adversarial environments, using fully autonomous, identical NAO robots.
• Guided the team's technical strategy by conducting a comprehensive literature review on multi-robot soccer, leading to a publication on language-driven world models [5]
• Designed and implemented the multi-agent coordination protocols for collaborative passing that formed a component of our team's winning strategy, applying concepts from my research [1]

Dean's List, Honors in Computer Science Program, Letters & Science Honors Program

University of Wisconsin-Madison

Valedictorian of 600, 4.6/4.0 GPA, International Baccalaureate Diploma, State Golden State Seal Merit Award, State Seal of Biliteracy, 4-year Scholar Athlete

Agoura High School

Peer Reviewer

Served as a peer reviewer upon nomination by the respective program committees:

• Neurips 2025 Workshop on Scaling Environments for Agents (SEA)
• Neurips 2025 Workshop on Aligning Reinforcement Learning Experimentalists and Theorists (ARLET)
• Neurips 2025 Workshop on Interpreting Cognition in Deep Learning Models (CogInterp)
• Neurips 2025 Workshop on Efficient Reasoning (ER)
• Neurips 2025 Workshop on Data on the Brain and Mind Findings (DBM)
• Neurips 2025 Workshop on Symmetry and Geometry in Neural Representations (NeurReps)

Journal Club Host and speaker

• Presented a critical analysis on the origins of intelligence, synthesizing concepts of scale-free cognition and developmental bioelectricity, and referencing work by Dr. Levin to discuss how higher-level agency evolves from the homeostatic, problem-solving capabilities of cellular collectives.
• Facilitated a post-talk discussion with researchers on the future of synthetic biological intelligence and the applications of the mechanisms presented.
• Invited to give future talks in recognition of the presentation's quality and the engaging discussion.

Co-Founder

Madison Machine Learning

• Co-organizing and growing a new community hub connecting machine learning students, faculty, and industry professionals.
• Co-leading weekly technical deep-dives and facilitating critical discussions on state-of-the-art papers in machine learning.

Executive Boardmember and Webmaster

AI Club

• Organized projects, designed a website, and managed resources to support students interested in artificial intelligence.

Vice President and Vice-Captain

Badger Ballroom Dance Team

• Organized and lead instructional workshops in collaboration with faculty and professionals across multiple universities and companies.

Executive Boardmember and Webmaster

Ballroom Association UW-Madison

• Designed and built a user-focused website to facilitate member communication.

Science Tutor

Agoura Highschool Science Honors Society

• Taught students advanced science concepts, raising understanding and exam scores.

Math Tutor

Agoura Highschool Math Honors Society

• Taught advanced math concepts to students, improving understanding and exam scores.

Lead Developer and Founder

Thunder Warrior Gaming

• Designed, built, self-published, and marketed the video game Thunder Warrior: Genesis.
• Achieved net profit.
• Engineered a custom game engine, multiplayer server, database, 3D models, and animations.