C9h-LLM: Jetson Thor Inference Platform • Collier Technologies LLC

• Architected the C9h-LLM Serving Platform: Engineered a high-performance headless IPC worker model utilizing Unix Domain Sockets and AnyEvent-based asynchronous dispatch to decouple HTTP gateway logic from high-latency GPU inference tasks.
• Gemini API Compatibility: Delivered a drop-in Gemini API replacement with support for NDJSON token-level streaming and real-time vision (VLM) image tensor processing.
• Successfully ported the Qwen 2.5 Coder family (1.5B, 7B) and DeepSeek-R1 (7B, 14B) to NVIDIA sm_86 (Ampere); optimized and deployed the 32B models on sm_110 (Blackwell/Thor).
• Developed the "Chipping Process" build strategy: Utilized atomic cache persistence to successfully compile 100GB+ models on memory-constrained edge hardware.
• Engineered multi-tier memory reclamation logic (`cudaMemPoolTrimTo`, `malloc_trim`) to stabilize TensorRT profiling and prevent OOM events during high-concurrency inference.
• Achieved 13.84 TPS sustained throughput on DeepSeek-32B with a verified stable 64k context window using NVFP4 quantization on Blackwell.
• Hardened aarch64 native build pipelines for Jetson Thor, resolving environment-specific library pathing and dependency conflicts for XS-based runtimes.
• Deployed concurrent model serving architectures using systemd user services and localized NVMe storage to eliminate I/O bottlenecks and maximize hardware utilization on Jetson Thor.

Graphics Platform & Visualization • Collier Technologies LLC • 2004 – 2019

• Codebase Maintainer: Primary developer and maintainer of GtkGLAreaSharp, providing foundational .NET bindings for OpenGL integration within the Gtk# and Mono ecosystems.
• Engineered Core Graphics Logic: Implemented low-level OpenGL object manipulation (rotation, translation, scaling) and a specialized Quake .mdl file loader for 3D asset ingestion.

Cloud & Big Data Hardening • Google • 2019 – Present

• Lead maintainer for Dataproc GPU Initialization Actions, orchestrating automated driver deployment and runtime optimizations for thousands of production nodes.
• Source-compiled and deployed PyTorch 2.4.1 and TorchVision for the sm_120 architecture (RTX 6000 Blackwell Server Edition), patching C++ source to resolve multi-digit CUDA versioning and parallelizing builds across 64 cores.
• Developed a comprehensive automated environment reproduction toolkit for Dataproc, enabling rapid setup of private VPC networks with Secure Web Proxy (SWP) egress and TLS interception.
• Engineered a robust Secure Boot signing pipeline for kernel modules on Debian and Rocky Linux, integrating with GCP Secret Manager for automated certificate lifecycle management.

Google, LLC. • 06/2019 - present • Technical Solutions Engineer (Big Data / AI)

Lead maintainer for Google Dataproc GPU Initialization Actions. Orchestrated the automated deployment of NVIDIA drivers and runtime optimizations for thousands of production nodes. Spearheaded a robust Secure Boot signing pipeline for kernel modules on Debian and Rocky Linux, integrating with GCP Secret Manager for automated certificate lifecycle management. Developed the automated environment reproduction toolkit, enabling rapid validation of complex proxy-egress and private cluster configurations for enterprise partners. Authored automated GPU validation suites for TensorFlow and PyTorch to verify driver integrity and Spark-on-GPU performance in restricted-network environments. Triaged complex failures across the distributed stack, including Spark/YARN application layers, GPU driver initialization, and distributed ML image generation pipelines using sm_120 hardware.

Spearhead software and systems design, development, and operations for diverse client environments. Manage all aspects of technical staffing, administration, and billing. Operating AS25819 since June 2020. Contributed a critical proxy bypass fix to the psf/requests library, resolving long-standing `no_proxy` handling issues for enterprise cloud integrations.

Shiftboard, Inc. • 05/2018 - 06/2019 • Remote Software Developer

Developed backend features for a cloud-based web application as part of a 5-person engineering team. Implemented production-ready fixes and automated verify-and-deploy integration pipelines. Engineered API service augmentations and verifiable verification frameworks for MySQL and PostgreSQL data layers.

The Linux Foundation • 02/2016 - 04/2018 • Release Engineer

Represented The Linux Foundation (LF) at technical steering committees and development forums. Orchestrated the development and deployment of Vagrant, Packer, Ansible, and Docker image generation tools. Managed request queues for large-scale open-source projects including FD.io, Zephyr Project, and IoTivity.

The Walt Disney Company • 11/2014 - 12/2015 • Senior Consultant

Lead Developer for a highly-scalable content distribution management portal. Automated customer requests to Tier-1 CDN providers (Akamai, Limelight, Level3) via asynchronous daemon architectures. Engineered a pricing calculator for internal compute/storage comparison with major cloud providers using NodeJS and AngularJS. Applied rigorous Test-Driven Development (TDD) using Java and Perl.

ZeroLag Communications, Inc. • 10/2013 - 07/2014 • Contract Engineer

Responsible for the operational health of data center networking equipment, including routers, firewalls, and load balancers. Automated repetitive network engineering tasks through custom role-based web interfaces. Managed deployment and security for Gentoo and Debian Linux firewall fleets.

Intel Labs • 07/2013 - 09/2013 • Senior Technical Consultant

Developed state-of-the-art provisioning infrastructure for arbitrarily large cluster environments using IPMI, PXE, and Cobbler. Implemented Software-Defined Networking (SDN) patterns for node-level tiered storage and network hardware orchestration.

Juniper Networks, Inc. • 03/2013 - 06/2013 • Technical Consultant

Managed the migration and support of legacy FreeBSD platforms requiring Active Directory authentication. Hardened unix-to-windows authentication using Kerberos, LDAP, and NTLM across the enterprise unix fleet.

F5 Networks, Inc. • 11/2010 - 08/2012 • Project Management Engineer

Architected IT operational administration systems for network appliance research teams. Developed wireless network optimization platforms and Java-based API integrations. Authored canonical deployment guides for high-availability RADIUS AAA load balancing.

Education

• 2009-2010: University of Washington, M.S. Program in Computational Linguistics
• 1999-2000: Olympic College, IT Concentration

Community Activities

• 1994-present: Technical support volunteer for various Internet support communities.
• 2002, 2018-present: Member, Mensa
• 2001-present: Technical Contributor to various Open Source projects (Mono, GNOME, MySQL)
• 2009-present: Amateur Radio Operator, KF7BMP (General Class). Secretary, Alphabet Amateur Radio Club (since 2024).
• 2009-present: Member, Association for Computing Machinery (ACM) (Active since 2025).
• 2010-2011, 2018-2019: Precinct Committee Officer, San Juan County

References

Rick Boucher (rboucher@sanjuanweb.com), Administrator, Orcas Online, Inc.
Jon Torrez (jon.torrez@gmail.com), Network Engineer
Bob Blair (massody@comcast.net), Project Management Engineer, F5 Networks, Inc.
Randal Schwartz (merlyn@stonehenge.com), Perl Expert, Stonehenge Consulting
Paul Horton (horton-p@aist.go.jp), AIST (Computational Biology Research Center)