Stephanee Lawson@Mary-J-Contrary
1
Events attended
1
Submissions made
United States
1 year of experience
About me
Some people get intimidated by the pace of AI; I just see it as another system to troubleshoot. As an alternative mom who’s always had a niche for tech, I specialize in rapid learning and "figure-it-out-ness." I’m the person you want on your team when things go sideways or when you need someone to bridge the gap between high-tech and real-world application. My aesthetic? Antique shops and tattoos. My work ethic? Learning your entire stack in an afternoon. Let’s build something cool and grab some good food afterward.
🤓 Latest Submissions
QATALYST SDS — AMD Edge Demo
QATALYST SDS is a transparent agent‑workflow framework designed to demonstrate that AI systems can operate reliably, inspectably, and autonomously on AMD hardware. It addresses a critical requirement: enabling operators to audit an AI’s decision‑making, especially when conditions shift or cloud infrastructure is unavailable. Core Architecture & Observability To move beyond opaque “black box” behavior, the demo exposes a traceable four‑stage control loop: The Agent — Generates the initial task response. The Critic — Performs real‑time evaluation and internal validation. The Scheduler — Determines the precise next step in the routing flow. The Ledger — Records every action and state change for full auditability. The system includes live telemetry controls—Substrate Temperature and Network Tension—allowing users to simulate hardware constraints or infrastructure stress. Adjusting these values shows exactly how the workflow adapts under load, making the system’s internal behavior fully visible. Hardware Strategy & Resilience To avoid cloud‑provisioning delays, QATALYST SDS was validated exclusively on local AMD hardware. This ensured a stable, reproducible demonstration with zero external dependencies. Proving the framework’s efficiency on edge‑class silicon establishes a clear benchmark that it will scale seamlessly to higher‑end AMD hardware. The “Safe‑Open” Philosophy The project is intentionally modular. The orchestration loop is safe, open, and designed for public inspection. More advanced logic modules remain private by design. This separation reflects a responsible development model—balancing structural transparency with necessary IP protection. Impact QATALYST SDS provides a straightforward, observable blueprint for making agentic systems reliable and accessible. It delivers the transparency and predictability required by engineers, operators, and stakeholders who must trust the behavior of the AI systems they deploy.
10 May 2026
