Author

Date of Award

2026

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Computer Science

Committee Chair

Tathagata Mukherjee

Committee Member

Letha Hughes Etzkorn

Committee Member

Jacob Hauenstein

Research Advisor

Tathagata Mukherjee

Subject(s)

Cloud computing, Database management, Multiagent systems, Artificial intelligence, Natural language processing (Computer science), Data privacy

Abstract

Cloud storage cost is an increasing concern for organizations, including the Commercial Satellite Data Acquisition (CSDA) program, which holds petabytes of satellite imagery in Amazon S3. How the data are stored and how often they are accessed determines the bill. At the petabyte scale, manually tracking the cloud state and arriving at valid cost-saving recommendations is a slow and non-scalable approach, and automating it with an agent could help. Given the privacy of these data, this thesis explores whether an Amazon-generated S3 inventory report, given as input to an agentic system built on a local large language model (LLM), can produce defensible, auditable cost-saving migration plans. The multi-agentic system is run on two real CSDA production buckets, SDX and Maxar, and produces recommendations that beat a status quo baseline, converge across model families, and improve in numerical accuracy through verifier-driven retries.

Available for download on Thursday, August 05, 2027

Share

COinS