Michael Gulden
Fractional Chief AI Officer · AI Systems Engineer · Founder
Mbgulden@gmail.com · (808) 724-1218 · Meridian, ID · michaelgulden.com · github.com/mbgulden · beyondsaas.ai · prismaticengine.com
Summary
AI engineer and founder with 13 years in technology (since 2013) spanning web development, UX/UI design, and full end-to-end project delivery — client discovery, design, build, and handoff. Working full-time in AI engineering since November 2025, when agentic AI became production-viable: designing and operating multi-agent systems in production, building LLM-powered products (API + MCP), and running a fractional Chief AI Officer practice that decomposes business processes and rebuilds them as agentic workflows — all on self-hosted AI infrastructure I built and operate.
Core competencies
- Agentic systems: task/agent decomposition, orchestration patterns, multi-agent development, context and memory strategy, long-horizon tasks
- Process reimagination: current-state mapping, eliminate / simplify / agentify analysis, workshop-based redesign, requirements translation
- Trust and control: human-in-the-loop controls, least-privilege permissions, evaluation criteria, validation of autonomous behavior
- LLM application building: prompt and context engineering, tool/function calling, MCP servers and connectors, API design and integration
- Playbooks and patterns: reusable reference flows, decomposition patterns, agent playbooks (open-source, pstack)
- Delivery: Python, JavaScript/TypeScript, SQL, APIs; local LLM inference (vLLM) on owned GPU infrastructure; in-VPC / hybrid deployment
- Facilitation: discovery with business owners, executive communication, mentoring engineers and agent operators
Experience
Founder / Fractional Chief AI Officer — BeyondSaaS.ai
AI implementation consultancy: in-VPC, cloud, and hybrid AI deployments for companies that need AI working inside their own operations.
- Run structured AI opportunity reviews and workshops with business owners: map current processes, identify automation candidates, and convert findings into buildable requirements and roadmaps.
- Client engagement: Eden Group (construction / custom home building) — entered through a paid AI Opportunity Review and progressed to a week-long AI workshop engagement to redesign company workflows.
- Design agentic solutions clients can adopt and scale: agent workflows, prompt/context strategies, integration requirements, and human-in-the-loop checkpoints.
Creator / Lead Engineer — Prismatic Engine (open source, AGPLv3)
Multi-agent orchestration engine (prismaticengine.com), designed and built end to end; runs a production fleet of AI agents that operate two live businesses daily.
- Architected agent decomposition, orchestration, and review systems: every agent output passes mechanical verification (tests, counts, evidence) before it lands — human-in-the-loop control by design.
- Led a measured fleet-reliability program: baselined 213 production agent runs (46% completion, 65% first-attempt success), then built the fixes — budget-checkpoint plugin (49-test suite), automated review sweeps, and playbook routing — against that baseline.
- Authored the pstack agent playbooks and the hermes-muse-bridge tooling (both public, MIT): reusable decomposition patterns and reference flows other operators run their own fleets with.
- Contributed kanban-stream, a work-streaming plugin merged upstream into NousResearch's hermes-agent (October 2026).
Creator / Lead Engineer — Human Design Engine
Commercial chart-computation API and MCP server (humandesignengine.com), designed, built, and operated as a product: full API + MCP surface for programmatic access, security-audited and hardened (spam/abuse controls, key-gated tiers).
Production AI Infrastructure — self-hosted (Meridian, ID)
- Multi-host Proxmox cluster on enterprise servers (Lenovo, Supermicro EPYC, Dell PowerEdge) with an RTX 3090 GPU pool.
- Local LLM inference with vLLM: tensor-parallel serving, FP8 KV cache, 262K-token context, speculative decoding, prefix caching.
- 40GbE networking with RDMA for low-latency GPU/cluster traffic; k3s (lightweight Kubernetes) for container orchestration.
- Secured remote access and segmentation via Tailscale; the same in-VPC / on-prem pattern deployed for consulting clients.
Founder — Sentinel IT Asset Disposal & Logistics (Meridian, ID)
Secure IT asset disposal for Treasure Valley businesses: pickup, documented chain of custody, certified data destruction, responsible recycling. Hardware lifecycle fluency across servers, storage, and networking fleets; vendor and logistics operations run end to end.
Founder / Operator — Active Oahu Tours (Kailua, HI — run remotely)
Kayak and e-bike tour company operated remotely from Idaho through systems and automation: booking platform operations, customer lifecycle campaigns, and AI-assisted back-office processes. The operating proof behind the consulting practice.
Web Developer / UX-UI Designer — Independent (2013 – present)
Front-end web development and UX/UI design, including contract work for Silicon Valley websites. Ran projects end to end — client conversation, design, build, handoff — rather than a single stage of the pipeline. Adobe Creative Suite.
Education
Bachelor of Arts, Graphic Design — Brigham Young University–Hawaii, 2013
Links
- github.com/mbgulden — prismatic-engine, hermes-muse-bridge (v1.5.0), muse-pstack (v1.1.0), hd-platform
- beyondsaas.ai · prismaticengine.com · humandesignengine.com · sentinelitad.com