Tucson, Arizona / Independent research & engineering
About TechSoft Systems
An independent technology practice and AI research lab, built on the experience of making real systems work.
Founded in 2012 by Johnathan W. Carroll, TechSoft connects a background in business IT, infrastructure, and recovery with hands-on research into private AI systems.
The company
From business recovery to applied AI.
TechSoft began as a consulting practice serving small and midsize businesses through systems engineering, network design, and technical support. An early engagement with TBM Equities started with a ransomware recovery: Johnathan restored affected services and returned business operations to service within 24 hours. That work grew into a multi-year infrastructure and IT leadership relationship.
Today, TechSoft primarily serves as Johnathan's independent applied AI R&D practice, alongside his full-time Tribal government role. Its dedicated AI lab began in January 2025 and has grown into an environment for building, testing, and evaluating inference platforms, agent systems, retrieval, and AI security.
Inside the lab
Research grounded in engineering.
The focus extends beyond running a model: how it is served, what information it can use, how its output is evaluated, and what happens when something fails.
Infrastructure & inference
Custom GPU systems, Linux and CUDA, NVFP4 conversion with NVIDIA Model Optimizer, and vLLM deployment. Workload-specific tuning supported by monitoring, reproducible tests, and recovery procedures.
PIPER intent routing, NVIDIA AI-Q research workflows, and human-reviewed memory systems. Investigating how context, evidence, and task boundaries survive the work performed by models and agents.
AI security testing, workload isolation, data sensitivity, tool-access boundaries, and governed use of external models. Research that informs architecture, policy, and training plans.
This is independent, AI-assisted research and engineering. Lab implementations, evaluated results, and future concepts are distinguished in the accompanying research records; they are not presented as completed commercial AI deployments.
Founder, owner & operator
Johnathan W. Carroll
Infrastructure, Security & AI Platform Engineering Leader
Johnathan brings 25+ years of experience across entrepreneurship, SMB and enterprise IT, ISP operations, utilities, and government. His background spans an earlier web-development business, third-party Microsoft support, leadership of a 52-person ISP operations organization, international hospitality network engineering, and physical-security integration at Tucson Electric Power.
Alongside TechSoft, he serves as IT Infrastructure Manager for the Pascua Yaqui Tribe Government. Across his Tribal tenure, infrastructure modernization and consolidation delivered approximately $250,000 in annual operating savings and a 50% reduction in physical footprint, alongside work in cybersecurity, disaster recovery, and data-center architecture.
His independent research informs the Tribe's emerging AI architecture and governance program, where policy review, technical development, and controlled-pilot planning remain in progress.
The approach
Understand intent. Verify evidence. Design for recovery.
The same questions connect infrastructure operations and AI research: What is the system supposed to do? What evidence shows it worked? What can be recovered when it does not?
TechSoft's work emphasizes observable systems, reproducible evaluation, preserved state, explicit trust boundaries, and human review. Capabilities matter, but so do their limits and the people who depend on them.
That perspective also shaped Defiantly Not Regex, a visual thesis from the lab about intent, imperfect communication, and the difference between matching a pattern and understanding a person.
