Ashton Shehan

At Walmart, I build and maintain internal software for the Alarm Central Service Team, turning frontline alarm-operations problems into practical tools. My deployed reporting and import tools save an estimated 25 staff-hours per week compared with prior manual workflows.

I built the Point Assignments Importer to reduce manual conversion work from 5–60 minutes to seconds, and developed Panel Validator and MASkr to improve configuration checks, auditability, triage, and rollback for alarm-code removal cases.

Through Actual Intelligence LLC, I create independent systems and security R&D projects including canary, an offline GGUF security auditor, and Sleeper, a GPU inference engine for large MoE checkpoints beyond available GPU memory. Sleeper reached 14.5 tokens per second on Qwen3-30B using 10.2 GB of RTX 4070 VRAM.

I’m a self-taught software engineer who combines alarm-operations knowledge with Python, CUDA, security analysis, automation, and ML systems work.