Chandler Weeks

About

Jason Chandler Weeks

Knoxville, TN

Graduate Research Assistant

Computer Science

Research

I am a graduate researcher in machine learning, big data analytics, high performance computing, and geospatial analytics. I am pursuing my Master's degree in Computer Science at the University of Tennessee, Knoxville, where I am a research assistant at the Global Computing Lab. My current research focuses on developing deep learning models for soil-moisture inference, and extracting explainable and interpretable insights from said models.

Engineering

I am also interested in software engineering, and have experience with a variety of programming languages and frameworks. I have a strong background in Python, and have experience with C++, R, and JavaScript. I am also proficient in using tools such as Git, Docker, and Kubernetes for software development and deployment.

Personal

In my free time, I enjoy running, hiking, and brewing coffee. I am super passionate about distance running, and have completed the Knoxville Marathon in 2026. I enjoy training, and hope one day to be Boston Bound. I also enjoy hiking, often visiting new trails to hike and camp with my girlfriend in the Smokies.

Projects

Machine Learning

SOMOSPIE

SOil MOisture SPatial Inference Engine

PyTorch · GDAL · Python · IBM Prithvi-EO-2.0 · HPC Systems

GitHub

Research Question

Can geospatial foundation models produce accurate, high-resolution soil-moisture predictions from coarse satellite-derived inputs that outperform traditional downscaling methods?

SOMOSPIE downscales coarse, satellite-derived soil-moisture estimates into high-resolution predictions.

I fine-tuned IBM's Prithvi-EO-2.0 with a convolutional decoder over HLS satellite imagery and 15 terrain parameters, taking 27 km² ESA-CCI inputs down to 10 m² over Alabama — a 2700× resolution increase that preserves the underlying spatial structure. The result is far more spatially expressive than the classical SOMOSPIE-ML (Random Forest) baseline; current work focuses on explainability and trust, validating that those predictions are physically meaningful and not just visually plausible.

Resolution

27 km² →10 m²

Upscale

2700×

RMSE

0.014

Before · 27 km² ESA-CCI

Coarse 27km ESA-CCI soil moisture over Alabama

After · 10 m² Prithvi

High-resolution 10m soil moisture prediction over Alabama

Traffic Signal Optimization Agents

SUMO · Reinforcement Learning · Python

GitHub

Trained reinforcement learning agents to control traffic signals in a SUMO simulation of a full city road network, replacing conventional fixed-timer controllers with policies that adapt to real-time vehicle flow. Each intersection's agent learns from local queue length and waiting time to decide when to switch phases, coordinating implicitly with neighboring intersections to keep traffic moving rather than running on a static schedule.

RL agent controlling a single intersection in SUMO

Agent-controlled intersection adapting phase timing to live demand.

City-scale SUMO simulation with vehicles distributed across the road network

City-scale deployment across the full SUMO road network.

Travel Time

-13%

Collisions

-25%

Avg. Queue

230 → 55veh

Systems & Architecture

Memory Hierarchy Simulator

C++ · Virtual Memory · TLB · Multi-Level Caches · COSC 530 @ UTK

GitHub

A simulator of the full memory access path. Every memory reference is walked through a data TLB, a page table, an L1 data cache, and an L2 cache before falling through to main memory and disk — with set count, associativity, line size, page size, and write policy all configurable per level.

TLBPage TableL1 CacheL2 CacheMain MemoryDisk

Per-access trace table showing TLB, page table, and cache results for each virtual address

Per-access trace: each virtual address resolved to its VPN, page offset, TLB / page-table / cache tags and indices, and hit or miss at every level.

CPU Scheduler Simulator

Python · pandas · matplotlib · NumPy

GitHub

A notebook that simulates eight classic CPU scheduling policies from a CSV of processes, renders the resulting execution timeline as a Gantt chart, and reports per-process response, waiting, and turnaround times alongside the averages.

FCFSSJFSRTFRound RobinPriorityPriority (Preemptive)Multilevel QueueMLFQ

Gantt chart for Shortest Remaining Time First preemptive scheduling

Shortest Remaining Time First (preemptive): short jobs cut ahead, deferring the longest process (P3) to the end.

Gantt chart for Multilevel Queue scheduling

Multilevel Queue (Q0 = FCFS, Q1 = RR q=3, Q2 = FCFS): higher-priority queues preempt lower ones, slicing P1 across the timeline.

Experience

Aug 2025 - Present

Knoxville, TN

PyTorchGDALHPCGeospatial

Graduate Research Assistant

Global Computing Laboratory

Enhanced a soil moisture inference pipeline across North America, increasing spatial resolution from 27 km² to 30 m² by fine-tuning geospatial foundation models and integrating a forecasting algorithm.
Designed scalable data preprocessing pipelines to remove cloud contamination and optimize terabytes of geospatial satellite imagery datasets for high-accuracy soil moisture forecasting.
Authored and deployed technical documentation for existing soil moisture inference tools to support development and usability for scientists.

Feb 2024 - May 2025

Starkville, MS

Scikit-LearnPythonUSDAVisualization

Undergraduate Research Assistant

Mississippi State Dept. of Computer Science

Developed a machine learning algorithm to predict changes in Vesicular Stomatitis Virus cases per month for use as an early-warning strategy for disease spread.
Refined preprocessing steps to cluster results to eco-regions, maximizing interpretability and algorithm efficiency without compromising accuracy.
Collaborated with USDA scientists to design a scientific data visualization tool, improving researchers' ability to analyze and communicate results.

May 2023 - Aug 2023

Ridgeland, MS

SQL ServerPower BIPython Automation

Software Developer Intern · Tools & Automation

C-Spire

Modified and maintained a Microsoft SQL Server database to address evolving software requirements.
Created comprehensive documentation for software products developed by the Tools & Automation team, increasing usability for external teams.
Developed a dynamic KPI reporting model leveraging Power BI to automate data updates, visualize performance trends, and support decision making.

Skills

Research Interests

Machine LearningBig Data AnalyticsHPCGeospatial AI

Languages

PythonRC / C++JavaScriptHTML / CSSSQL

Libraries / Tools

PyTorchScikit-LearnPandasNumPyGDALMatplotlibPyTestReact

Environments

LinuxGitDockerJupyterPower BIExcelGoogle Earth EngineQGIS

Education

University of Tennessee, Knoxville

M.S. Computer Science

GPA 3.58 / 4.00 · Knoxville, TN

Expected May 2027

Mississippi State University

B.S. Software Engineering

GPA 4.00 / 4.00 · Starkville, MS

May 2025

Relevant Coursework

Computer Architecture
Database Management Systems
Artificial Intelligence
Machine Learning
Deep Learning
Natural Language Processing
Computer Vision
Advanced Algorithms

Contact

I'd love to talk about my research or machine learning work. Feel free to reach out via email.

[email protected]

GitHub: github.com/ChandlerWeeks
LinkedIn: linkedin.com/in/chandler-weeks
Location: Knoxville, TN
Resume: Weeks_Jason.pdf · 2 pp