Willard Hucks

Principal Engineer · Enterprise Systems Engineering

Location: McMinnville, Oregon
Email: 127.0.0.1@willardhucks.com
GitHub: github.com/trydydd

Summary

"Principal engineer with 14 years across infrastructure, DevOps, and platform architecture, now leading applied AI at enterprise scale: local LLM inference (vLLM), agentic and MCP-based systems, governed RAG, and AI cost optimization. Sets direction for model selection, evaluation, and responsible AI consumption while staying hands-on; track record of architecting provider-agnostic platforms and multiplying the teams around them."

Key Highlights

"Designed MCP-based developer tooling that cut code analysis costs ~70% by feeding models precise, governed context instead of bulk source."

"Architected a schema-as-guardrails platform that lets citizen developers generate code safely to company specification, enforcing secure tool and data access through approved containers and reusable patterns instead of one-off solutions."

Experience

Principal Engineer, Enterprise Systems Engineering
Nov 2021 – Present | Nike, Inc. | Beaverton, OR
Senior Engineering Manager
Mar 2018 – Nov 2021 | Nike, Inc. | Beaverton, OR
Lead DevOps Engineer
Apr 2017 – Mar 2018 | Nike, Inc. | Beaverton, OR
DevOps Engineer
Aug 2015 – Apr 2017 | Nike, Inc. | Beaverton, OR
Service Virtualization Engineer
Jan 2015 – Aug 2015 | Nike, Inc. | Beaverton, OR
Systems Administrator
Mar 2014 – Jan 2015 | Performance Health Technology | Salem, OR
Helpdesk Manager
Oct 2012 – Mar 2014 | Allegheny Technologies Incorporated | Albany, OR
Armory Chief
Jan 2009 – Oct 2012 | United States Marine Corps | Jacksonville, NC

Projects

Synaptic Drift: governed context retrieval for LLM agents

Status: open source · in active development

  • Retrieval-augmented generation (RAG) system serving verifiable, governed context to agents over MCP with schema-enforced tool contracts; ~80% token reduction on documentation/context retrieval in testing to date.
  • Validated by end-to-end task success: building working functionality against live, unseen library docs using a 9B model, so retrieval quality couldn't be masked by model capability. Sub-21ms P95; cryptographic content verification and lifecycle governance.

View Repository

Vamp: schema-as-guardrails platform for citizen developers

Status: personal · in active development

  • Lets citizen developers write and ship code to company specification safely, enforcing secure tool and data access through schema-defined guardrails instead of manual review.

Skills

Applied AI

Large language models (LLMs) Agentic and multi-agent systems Model Context Protocol (MCP) Retrieval-augmented generation (RAG) LLM inference serving (vLLM) Model evaluation & experimentation Prompt & guardrail design Cost optimization & inference efficiency Responsible AI & model governance MLOps Model pretraining

Architecture & Leadership

Enterprise AI architecture Platform architecture Build-vs-buy analysis Technical direction & standards Mentorship Cross-functional collaboration Risk management Agile

Languages

Python Bash PowerShell

Platforms & Infrastructure

Azure AWS Docker Kubernetes Datacenter Nutanix Ansible Jenkins Git/BitBucket CI/CD Infrastructure-as-Code

Education & Honors

Generated by Self-hosted (vllm) · google/gemma-4-E2B-it