Adrian Lozada
M.S. Computational Science & Engineering @ Georgia Tech
I build high-performance computing systems, parallel algorithms, and intelligent software. My work spans HPC, LLM tooling, distributed systems, robotics, and machine learning.
Experience
Open Source Software Engineer Intern
May 2026 – Aug 2026Center for Scientific Software Engineering (CSSE), Georgia Tech — Atlanta, GA
- Prototyped an experimental LLM extractor that identifies case-law citations in legal briefs, using Mellea requirements and instruct-validate-repair sampling over backend-agnostic Ollama/LiteLLM endpoints.
- Raised detection recall from 22.2% to 83.3% and F1 from 34.8 to 85.7 on IBM Granite 4.1 (3–30B) models, benchmarked against the rule-based eyecite extractor on a frozen citation benchmark.
Software Engineer Intern
May 2025 – Aug 2025Orion Defense Solutions L.L.C. — Remote
- Built and deployed a Python/FastAPI automation platform integrating LangChain and SQL-backed internal systems, reducing project cycle time by ~8 hours.
- Containerized platform microservices using Docker and Kubernetes, enabling rolling updates and eliminating ~7 hours of manual setup per week.
Undergraduate Researcher
Jan 2024 – May 2025Reality, Autonomy, Robot Experience (RARE) Lab, USF — Tampa, FL
- Published two first-author papers at ACM/IEEE HRI 2025 (25% acceptance rate) on autonomous fog-screen communication systems.
- Developed ROS nodes in C++ to synchronize Fetch robot behavior, achieving reliable distributed actuation.
Projects
High-Performance Parallel Computing
- Parallelized dense matrix kernels using OpenMP work-sharing and loop-collapse strategies in C++, achieving near-linear 16x speedup at 16 threads on GT's PACE/ICE cluster.
- Accelerated a heat equation PDE solver on an NVIDIA H100 via CUDA shared memory tiling and OpenMP GPU offloading, achieving up to 83.9x and 17x speedup respectively over single-threaded CPU baselines.
Graph Neural Network-Based Fraud Detection
- Developed a Graph Neural Network in PyTorch to classify fraudulent financial transactions, achieving 91% F1 score on real-world datasets.
Autonomous Ground Vehicle with Remote Monitoring System
- Led a team of 7 to design and build an autonomous vehicle on Raspberry Pi, mentoring peers in Linux development and conducting code reviews to deliver a working prototype.
- Designed and deployed a Flask backend with SQLAlchemy to expose REST endpoints for robot status, logs, and commands.
Education
Georgia Institute of Technology
M.S. Computational Science and Engineering
Coursework: High Performance Parallel Computing, Numerical Linear Algebra, Modeling and Simulation
University of South Florida
B.S. Computer Engineering, summa cum laude
Honors: FLIT-GAP Scholar, Bright Futures Scholar
Coursework: Analysis and Design of Algorithms, Data Structures, Social Networks, Computer Architecture
Skills
Languages
Parallel / HPC
Tools
Libraries
Awards
UR2PhD Technical Conference Travel Award
Computing Research Association (CRA)
Tampa Conference Presentation Grant Program
CPGP
NSF Distributed Research Experiences for Undergraduates
DREU
Florida Bright Futures Scholarship
100% of Tuition