25 years building systems that don't go down.
Now building systems that think.
Progressive roles in DevOps, systems administration, network engineering, and infrastructure across Vets First Choice, EVO Payments, MEMIC, and others — building the Linux, networking, virtualization, and storage foundations that underpin modern platform and AI infrastructure work.
Self-directed research conducted alongside professional work, 2024–present. Private GPU hardware, production-pattern systems, real evaluations — not proofs of concept.
Dual-node cluster serving 30B–397B parameter models (Nemotron, Qwen, Cascade). SRE discipline applied to inference: quantitative benchmarking, config management, Prometheus/Grafana observability, custom vLLM fork for hardware-specific optimizations.
Production multi-agent system with agents for infrastructure ops, automated evaluation, project tracking, and workflow automation. Integrated with real operational tooling — not a demo.
Comprehensive eval combining industry-standard benchmarks (via lm-evaluation-harness) with custom agentic testing: 16 tool-calling tests, text quality assessment, multi-turn reasoning scenarios. InfluxDB + Grafana dashboards.
Executed preference alignment and parameter-efficient fine-tuning on hybrid architectures. Full-cycle pre/post evaluation for catastrophic forgetting detection. Developed internal guidelines for small-active-parameter MoE models.
Semantic search pipelines providing persistent contextual memory across agent sessions. Multi-phase AI pipelines with swappable LLM backends.
An AI assistant trained on my background is coming soon.
chris@whyland.net