Open to opportunities

Gavin Morris

Computer Science student at Louisiana State University building production systems across GPU kernels, LLM inference, AI platforms, and cloud backends — from Triton kernels and vLLM contributions to Spacebourne, the ATHENA Lab, enterprise healthcare, and founder-led client delivery.

View projects My experience
gmorriscs@gmail.com (504) 222-8798 LinkedIn gavinmorrisportfolio.com
Introduction

About me

I'm a Computer Science student at LSU with a minor in Mathematics, passionate about machine learning, systems performance, and shipping software that solves real problems — from writing Triton inference kernels on A100s and contributing to vLLM, to AI workflow engineering at Spacebourne (Delta-V Scholar Capstone Award), GPU-accelerated diffusion research on LONI HPC at the ATHENA Lab, and founder-led delivery at Deltara Tech, where I built production systems for 100+ users across Louisiana businesses.

Education

Louisiana State University

Bachelor of Science, Computer Science

Minor: Mathematics

Baton Rouge, LA May 2027 GPA: 3.74
0
Roles & internships
0
Featured projects
0
Certifications
0
Campus organizations
Career

Experience

Hands-on engineering across AI platforms, ML research, healthcare, GIS, and founder-led client delivery.

June 2026 – September 2026

Software Engineer Intern

Spacebourne

Delta-V Scholar Award
  • Received the Summer 2026 Delta-V Scholar Capstone Award.
  • Developed 5+ production-ready features adopted across the CLARITY platform, improving user productivity and expanding system capabilities.
  • Optimized backend processes and automated AI workflows, reducing AI processing time by 20% and manual review time by 35%.

June 2026 – September 2026

Undergraduate AI Research Assistant

LSU ATHENA Lab

  • Engineered PyTorch diffusion pipelines executed on NVIDIA A100 GPUs on LONI's QB-4 supercomputing cluster, implementing configurable denoising schedules, latent upsampling strategies, and multi-stage image generation workflows.
  • Built automated benchmarking infrastructure for large-scale GPU experiments on HPC infrastructure, orchestrating Slurm batch jobs and logging generation outputs, GPU runtime performance, and evaluation metrics across hundreds of test configurations.
  • Profiled GPU inference tradeoffs between speed, image quality, and text-image alignment using Stable Diffusion and Hugging Face Diffusers, characterizing latency and memory behavior across pipeline configurations.

January 2026 – May 2026

Software Engineer Intern

Blue Cross Blue Shield of Louisiana

  • Developed backend features using C# and SQL to manage client documents, payments, and records within cloud-based systems, ensuring accurate storage and retrieval.
  • Built and optimized database queries and workflows to improve data consistency and performance across patient-facing applications.
  • Collaborated with engineers and stakeholders to debug issues, deploy updates, and enhance system reliability in a production environment.

May 2025 – July 2026

Founder & Software Engineer

Deltara Tech LLC

  • Generated $18,000+ in revenue building custom software solutions for 6 Louisiana businesses.
  • Delivered production web, mobile, and automation systems used by 100+ end users across HVAC, roofing, and home-service industries.
  • Managed client acquisition, requirements gathering, development, deployment, and ongoing support.

January 2025 – May 2025

Software Developer Intern

LSU CARTS

  • Collaborated with a team of two to develop GIS applications using Python and Blazor.
  • Developed applications that enhanced the efficiency of GIS analysts, increasing their productivity by up to 25%.
  • Applied problem-solving techniques and programming expertise to deliver functional and efficient GIS tools that support research and transportation safety initiatives.
Portfolio

Projects

From ML inference systems to full-stack and mobile apps — filter by focus area.

2026 ML & AI

Triton Inference Kernels

  • Wrote Transformer inference kernels (softmax, RMSNorm, tiled GEMM) from scratch in Triton, sustaining 82% of peak HBM bandwidth on A100 (72% on T4) and 96% of cuBLAS FP32 throughput with a single unchanged source across both GPU generations; validated against PyTorch on non-power-of-two shapes.
  • Diagnosed a 20× FP16 GEMM slowdown down to the instruction level by dumping compiled PTX: Triton emitted zero tensor-core (mma) instructions on sm_75 vs. 8 on sm_80, confirming a compiler codegen gap (not kernel code) that produced a 30× jump (2.4 → 73 TFLOP/s) on identical source across GPU generations.
  • Built a cross-vendor benchmark suite (roofline analysis, median-of-N timing, FP32/FP16) spanning T4, A100, and AMD MI300X cloud instances; documented Ampere's silent TF32 default breaking bit-level FP32 reproducibility and a full failure taxonomy of ROCm environment provisioning in a public friction log.
View on GitHub
2026 Open Source

vLLM

  • Merged a documentation fix into vLLM, the leading open-source LLM inference engine, tracing doc–code mismatches through the model registry and test suite.
  • Approved through core-maintainer review in a single cycle.
View pull request
2026 ML & AI

Inference Kernel Optimization Lab

  • Implemented Transformer inference kernels (GEMM, Softmax, LayerNorm) in a three-tier progression — pure Python baseline, NumPy/BLAS, and JIT-compiled Numba — achieving a ~300× speedup over baseline through cache-block tiling, matrix-transpose memory layout optimization, 4-way loop unrolling, and thread-level parallelism.
  • Applied roofline model analysis to classify kernels as compute-bound (GEMM) vs memory-bound (Softmax, LayerNorm) by arithmetic intensity and quantified the remaining gap to hand-tuned BLAS, attributing it to SIMD auto-vectorization limits.
  • Engineered a rigorous benchmarking harness with median-based timing, JIT warmup control, single-thread isolation for fair kernel comparison, and block-size sweeps, producing reproducible CSV results and performance plots.
View on GitHub
2026 ML & AI

ML Inference Scheduler & Performance Profiler

  • Built an SLO-aware ML inference scheduler modeling mixed workloads with micro-batching, priority queuing, and GPU-style execution lanes under memory-bandwidth constraints, benchmarking 4 scheduling policies head-to-head.
  • Quantified latency, throughput, p99 tail latency, and SLO violation rates via trace-based profiling, characterizing queue contention and latency/throughput tradeoffs across batching strategies.
View on GitHub
2026 Full-Stack

CoHabit

  • Developed a JavaFX (Java 17) desktop application for shared-living coordination — room setup, chore tracking, expense balancing — with layered model/core/UI architecture, Maven tooling, and JUnit 5 coverage of core manager logic.
  • Engineered a dual-persistence layer integrating Firebase Firestore via REST with transparent local JSON fallback (Jackson), preserving full functionality offline or on cloud failure.
View on GitHub
November 2025 Mobile

Coach AI

  • Developed an AI-powered iOS platform combining LLMs, real-time health data, task management, and location-aware automation for personalized coaching and workout planning.
  • Integrated Llama 3.1 via Groq API into a context-aware pipeline aggregating HealthKit metrics, goals, tasks, and calendar events for automated workout plans and task updates.
  • Designed a scalable cloud architecture using Firebase Authentication and Firestore with dual-layer persistence, secure data isolation, and geofence-triggered automation via CoreLocation.
View on GitHub
May 2025 Mobile

Inventory Pro Tracker

  • Developed and launched a full-stack inventory management system featuring a Swift UI iOS app and responsive web interface (HTML, CSS, JavaScript), now available on the App Store.
  • Designed and deployed a scalable backend using Railway, enabling real-time tracking of inventory across multiple suppliers and locations.
  • Built intuitive UI/UX for businesses to easily manage complex supply chains with multi-supplier support, item categorization, and usage history visibility.
View on GitHub
Expertise

Technical skills

ML & AI

Triton CUDA (PTX) ROCm GPU kernel optimization LLM inference Transformers PyTorch

Systems & Performance

Roofline modeling Memory-bandwidth analysis Latency/p99 profiling Micro-batching Slurm/HPC

Infrastructure & Tools

Python SQL Docker Kubernetes Git/GitHub REST APIs Cloud GPU deployment (RunPod/Vast.ai)
Credentials

Certifications

Hover to reveal course IDs. Tap on mobile to flip.

Anthropic

AI Fluency: Framework & Foundations

Hover for ID
7hykoutix9vb
Wharton

Foundations & Applications of Financial Technology

Hover for ID
A9CABJKOSLQ3
IBM

Full Stack Developer Professional Certificate

Hover for ID
3UBKQ00NGCIA
IBM

Generative AI Engineering Professional Certificate

Hover for ID
EAMKPE6N4UHX
Community

Campus involvement

Leadership, service, and technical community engagement at LSU.

LSU Student Government

Assistant Director of Technology, Innovation, and Modernization

Driving technology initiatives and modernization efforts for student government operations and campus-facing solutions.

Student Technology Fee Committee

Student Representative

Representing students in decisions on how technology fee funding is allocated to campus technology resources.

LSU Companion Animal Alliance

Student Volunteer

Volunteering to support companion animal care and welfare initiatives on campus and in the community.

Artificial Intelligence Club

Student Member

Engaging with peers on AI research, projects, and emerging technologies in machine learning.

Tau Kappa Epsilon

Social Chairman — Executive Board

Serving as Social Chairman on the Executive Board, organizing events and fostering brotherhood within the fraternity.

Connect

Let's work together

I'm always open to internships, collaborations, and conversations about software and AI.

Email

gmorriscs@gmail.com

Phone

(504) 222-8798

LinkedIn

linkedin.com/in/gmorr32

Portfolio

gavinmorrisportfolio.com
Say hello