Sammi De Blas

LLM evaluation, llama.cpp, GGUF, prompt injection, MITRE ATT&CK, Splunk, Sysmon, Volatility3, malware analysis, PCAP, RTL-SDR, Python, Linux, log pipelines, Obsidian

Hardening local LLMs
Adversarial prompt batteries mapped to MITRE ATT&CK, prompt injection testing, and monitoring what each process sends to AI providers.

A local-first AI lab: benchmarking local LLMs for censorship and rejection rate, adversarial prompts mapped to MITRE ATT&CK, and an automation bus that runs everything on my own hardware.

Reviewing DFIR or malware write-ups, and discussing how a local LLM stack should be tested before it touches production.