Machine Learning Researcher (Fellow) · ORAU / AEOP · San Antonio, Texas
I turn research questions into working AI systems .
I design learning algorithms, train and evaluate models, build research infrastructure, and lead projects across RLHF, reinforcement learning, compact language models, intelligent agents, and autonomous systems.
AAAI 2024
ICML 2025 Workshop · Oral
NeurIPS 2025 ARLET
Research → Systems
End-to-end
Machine Learning
Machine Learning
RLHF
Small LMs
Agents
Evaluation
RL
End-to-end capability
From a new idea to evidence, software, and communication.
Research & algorithm design
Formulate research questions, design learning objectives, identify assumptions, and turn conceptual ideas into testable methods.
RLHF
Reward learning
Reinforcement learning
Model training & implementation
Build reproducible pipelines for reward models, policy optimization, compact language models, agents, and custom environments.
PyTorch
Hugging Face
MLX
Evaluation & experimentation
Design controlled experiments, diagnose failure modes, compare alternatives, and evaluate performance, reliability, and generalization.
Benchmarking
Calibration
Human studies
Leadership & communication
Lead projects through publication, coordinate technical work, mentor researchers, and communicate results to technical and executive audiences.
Project leadership
Mentorship
Presentations
Selected research
Questioning assumptions about feedback, scale, and learned behavior.
AAAI 2024
RLHF · Reinforcement Learning
Rating-Based Reinforcement Learning
A direct-rating alternative to preference-based RLHF that learns reward functions from evaluations of individual trajectories rather than pairwise comparisons.
My contribution: Helped originate the research direction and led algorithm design, implementation, experimental planning, analysis, and publication.
Human feedback
Reward modeling
Human study
ICML 2025 Workshop · Oral
Small models
Too Big to Think
Controlled arithmetic and factual-recall experiments reveal a capacity-dependent shift between extrapolation and memorization.
My contribution: Advised on research framing, interpretation of results, and communication of the central findings.
ICML 2025 Workshop
Reward modeling
Multi-Task Reward Learning
Combines classification and regression objectives with learned uncertainty weights to improve reward learning from limited rating data.
My contribution: Contributed to problem formulation, analysis of results, research direction, and manuscript development.
Working systems
Projects that turn research ideas into usable implementations.
Research codebase
Rating-Based Reinforcement Learning
Official implementation of a direct-rating alternative to preference-based reward learning and policy optimization.
Role: Research lead, algorithm design, core implementation, and evaluation.
PyTorch
RLHF
PPO
Open repository
↗
Benchmark & environments
Atari-GPT
Benchmarking multimodal language models as low-level policies in interactive Atari environments.
Role: Research direction, code, experiments, analysis, and writing.
LLMs
Agents
Evaluation
Open repository
↗
End-to-end training
Small-LLM RLHF Pipeline
Reward modeling, parameter-efficient fine-tuning, and RLOO policy optimization for compact language models.
Role: Independent end-to-end implementation and evaluation.
TRL
PEFT
LoRA
Open repository
↗
Agent architecture
Minimal LLM Agent Framework
A compact ReAct-style framework with pluggable tools, reflection, retries, structured traces, and SQLite-backed memory.
Role: Independent architecture and implementation.
Agents
Memory
Python
View project details
→
On-device inference
ASCII Breakout MLX
A compact Llama-based agent running an interactive ASCII game efficiently on Apple Silicon.
Role: Independent implementation and on-device evaluation.
Apple MLX
Llama
Edge AI
Open repository
↗
Training from scratch
RL-Driven nanoGPT
A GPT-style model trained from scratch with PPO to control an ASCII Breakout environment.
Role: Independent model, environment, and training implementation.
nanoGPT
PPO
PyTorch
Open repository
↗
Research journey
From electrical engineering to leading end-to-end AI research projects.
My work began with reinforcement learning as an undergraduate researcher and grew into a multi-year program spanning human-subject studies, algorithm design, model training, benchmarking, open-source software, publication, and mentorship.
See full experience2023—Present
Machine Learning Researcher (Fellow) · ORAU / AEOP
Leading and contributing to research in RLHF, compact language models, benchmarking, agent systems, and human-guided reinforcement learning.
2022—2023
Graduate Research Assistant · UTSA
Advanced Rating-Based Reinforcement Learning through algorithm development, human-subject research, controlled experiments, and publication.
2021—2022
Undergraduate research & engineering
Helped establish the RbRL research direction, led student implementation work, and developed reinforcement-learning systems for control and guidance problems.
Impact & recognition
Research impact beyond publications.
Invited & selected talks
Presented research for DEVCOM VIP Day, the ARL Technical Advisory Board, and AEOP/ORAU audiences.
Research mentorship
Mentored 7+ students while coordinating experiments, implementation, technical decisions, and research direction.
Peer-review service
Reviewer for IEEE Transactions on Cybernetics.
Recognition
Featured by UTSA as part of its first M.S. in Artificial Intelligence cohort; First Place at the UTSA Tech Symposium; recipient of the Outstanding Undergraduate Award for Excellence in STEM.
Read UTSA feature ↗Explore the work