Neil McGlohon

Neil McGlohon, Ph.D.

Research Scientist · Software and Systems Engineer

Parallel & Distributed Computing · Neuromorphic Computing

Résumé PDF Full CV w/ Pubs PDF

01 Education

Doctor of Philosophy
Computer Science
Rensselaer Polytechnic Institute
August 2021
Master of Science
Computer Science
Rensselaer Polytechnic Institute
May 2016
Bachelor of Science
Physics
University of Oklahoma
May 2014

02 Experience

Skills

  • AI/ML & Inference: PyTorch, TensorFlow, vLLM, Ray, agentic systems, neuromorphic hardware
  • Parallel & Distributed Computing: HPC, MPI, NCCL, SLURM, cloud-native architectures
  • Cloud & Infrastructure: Kubernetes, OpenShift, Helm, Docker, Podman, Grafana, Prometheus, Infiniband \& RoCE, NFS, Longhorn, MetalLB, AWS, IBM Cloud
  • Software Engineering: C/C++, Python, JavaScript/Node.js/Vue.js, GitOps (Git, Jenkins, ArgoCD), CMake, AMQP, S3, REST, SQL, MongoDB, asynchronous event-driven systems
Staff Research Scientist — IBM Research
Oct 2022 — Jul 2026
  • Led design and development of a purpose-built LLM inference engine for the IBM AIU NorthPole AI inference accelerator card, delivering state-of-the-art low latency and energy efficiency.
  • Led team of 5 engineers building NorthPole's performance-critical runtime stack.
  • Deployed rack-scale distributed AI systems of 288 NorthPole cards, generating over 500,000 tokens per second in aggregate while holding per-token latency as low as 1 millisecond.
  • Collaborated with model team to evaluate our quantized and retrained models using tools and benchmarks including EleutherAI's lm-eval and Hugging Face's OpenLLM leaderboards.
  • Designed, built, and deployed a cloud-native OpenShift/Kubernetes stack that exposes controlled API access to NorthPole LLM inference appliances and a chat front-end; integrated back-end into IBM watsonx, bringing NorthPole-accelerated inference to workloads across the platform.
  • Created NorthPole runtime client/server applications leveraging the YOLOv9 model to perform object detection, identification and tracking on real-time video streams, providing over 300 megapixels per-second of image processing consuming less than 60W of card power.
  • Guided broader team as subject-matter expert in parallel and distributed systems, containerization, orchestration, and LLM inference serving including vLLM plugin development and integration.
  • Represented NorthPole at business development meetings, client outreach and showcase events, presenting the platform to prospective industry and public-sector partners. Bridged scientific and business-focused teams, driving knowledge transfer and cross-team collaboration.
Research Scientist — Rensselaer Polytechnic Institute
Jul 2021 — Sep 2022
Center for Computational Innovations
  • Development and optimization of workloads on the HPC systems for internal and partner research.
  • Served on steering committee of an NSF-funded project-of-projects, mentoring participating graduate students as a subject-matter expert in HPC and AI systems.
  • External collaborator with IBM AI Hardware Center on research into distributed AI training benchmarking and simulation of composable systems.
  • Triaged research support and facilitation tickets for thousands of external users across three supercomputer systems.
Software Engineer — Rensselaer Polytechnic Institute
Feb 2020 — Jun 2021
Center for Computational Innovations
  • Maintained the open-source Rensselaer Optimistic Simulation System (ROSS) project, a parallel-discrete-event-simulation (PDES) engine capable of providing strong-scaling linear speedup even when distributed across 2-million supercomputer CPU cores.
  • Maintained CODES, an open-source C/C++ framework for massively parallel simulation of high performance computing (HPC) communication networks, collaborating with researchers at Argonne and Sandia National Laboratories. CODES is built on top of the ROSS parallel discrete event simulation (PDES) engine.
  • Collaborated with external research partners, writing specifications and architecting bespoke software solutions to capture bleeding-edge experimental results.
Software Engineering Intern — Cisco Meraki
Summer 2017
  • Researched and developed firmware for Meraki's MR platform access points.
  • Designed and built a lockless, thread-safe hashtable using RCU techniques, and upstreamed the contribution to Click, an open-source software-defined router project.
Scientific Service
2019-Present
Committee work
  • 2025 — ACM SIGSIM PADS, Program Committee Member
  • 2024 — ACM SIGSIM PADS, Program Committee Member
  • 2023 — ACM SIGSIM PADS, Program Committee Member
  • 2023 — ACM SIGSIM PADS, Organizing Committee — Publicity Chair
  • 2023 — IEEE IPDPS ScaDL Workshop, Program Committee Member
  • 2022 — ACM SIGSIM PADS, Organizing Committee — Ph.D. Colloquium Chair
  • 2022 — ACM SIGPLAN BID Workshop, Program Committee Member
  • 2022 — IEEE IPDPS ScaDL Workshop, Program Committee Member
  • 2020-2022 — NSF CAREERS Cyberteam Program, Steering Committee Member
Peer review
  • 2025 — ACM SIGSIM PADS'25
  • 2024 — ACM SIGSIM PADS'24
  • 2023 — ACM SIGSIM PADS'23
  • 2023 — IEEE IPDPS ScaDL Workshop
  • 2022 — ACM SIGPLAN BID Workshop
  • 2022 — IEEE IPDPS ScaDL Workshop
  • 2021 — IEEE Transactions on Parallel and Distributed Systems (TPDS)
  • 2021 — IEEE European Control Conference (ECC'21)
  • 2020 — IEEE Transactions on Parallel and Distributed Systems (TPDS)
  • 2019 — ACM SIGSIM PADS'19

03 Selected Projects

Homelab Infrastructure Platform
2022 — Present
  • Built and operate a 4-node ARM Kubernetes cluster running 30+ self-hosted services, each packaged as a Helm chart and continuously deployed from Git by ArgoCD. All secrets injected from 1Password operators to avoid any secret data being committed to Git.
  • Designed a walled-garden network for autonomous AI agents: a deny-by-default restricted VLAN, per-agent scoped Git identities, and a dedicated Jenkins instance running on-demand build pods in an isolated namespace, reducing the potential consequences and risk from errant agents.
  • Serve local LLM inference on RTX 5070 Ti, RTX Pro 2000, and Rockchip RK3588 NPU hosts, powering interactive and unattended coding agents entirely on-premises.

Homelab overview page

04 Publications

05 Talks

  • Guest Lecturer: Introduction to the Message Passing Interface (MPI) and High Performance Computing Systems
    Neil McGlohon.
    Parallel Computing, Rensselaer Polytechnic Institute — February 14-21, 2020
  • CODES-Tracer Tutorial: Enabling HPC Design Space Exploration via Discrete-Event Simulation
    Nikhil Jain and Neil McGlohon.
    HOT-Interconnects — August 2019, San Jose, USA
  • Introducing the Dragonfly Plus Interconnection Model to CODES
    Neil McGlohon and Misbah Mubarak.
    Summer of CODES Workshop — July 2018, Argonne National Laboratory, USA
  • Simulation and Visualization of Custom Neuromorphic Hardware using NeMo
    Mark Plagge, Neil McGlohon, Caitlin Ross, and Christopher D. Carothers.
    Neuromorphic Computing Symposium on Architectures, Models, and Applications — July 2017, Oak Ridge, USA · Extended abstract/presentation
  • Weak anti-localization in ultrathin Sb(111) films
    S. Cairns, N. McGlohon, C. Robison, J. Keay, C. K. Gaspe, K. S. Wickramasinghe, T. D. Mishima, M. B. Santos, and S. Q. Murphy.
    American Physical Society, APS March Meeting — March 2013, Baltimore, MD · Extended abstract/presentation

06 Other experience

Research Assistant — Rensselaer Polytechnic Institute
Fall 2014 — Jun 2021
Advisor: Christopher Carothers
  • Member of the Department of Energy's Exascale Compute Project (ECP), developing and evaluating models of prototypical exascale supercomputer communication network designs through simulation.
  • Fulfilled duties as a research assistant with the SuperNeuro research group, developing models for simulation of neuromorphic systems and associated software. Maintained ROSS and CODES open-source software projects.
  • Primary area of research: high performance/parallel computing. Other areas of interest: distributed systems, cloud computing, machine learning, neuromorphic computing and simulation.
Teaching Assistant, Computer Science 1 — Rensselaer Polytechnic Institute
Fall 2014
Professor: Sibel Adali
  • Supported the instructor of an introductory Python course.
  • Led lab sessions and discussions, held office hours, and proctored and graded exams/assignments.
Research Assistant — University of Oklahoma
May 2012 — May 2014
Advisor: Sheena Murphy
  • Explored electrical transport properties of quantum-confined antimony measured at cryogenic temperatures.
  • Created laboratory process for measuring differential conductance of a material, allowing for a deeper understanding of the interface between a topological insulator and a superconductor.
National Science Foundation REU — University of Oklahoma
Summer 2012, 2013
Advisor: Sheena Murphy
  • Condensed matter research within NSF Research Experience for Undergraduates program including first observations of weak anti-localization in quantum-confined antimony films.
  • Completed a metal-workshop training course to qualify for use of the student shop, fabricating parts for laboratory equipment.
Senior capstone research project
Aug 2013 — May 2024
  • Final year of undergraduate physics work at the University of Oklahoma, overseen by faculty members Sheena Murphy and Matthew Johnson — continuing work from two summer REUs and coursework research. Presentations given and critiqued by advisors and students throughout.
Editor of Photography, Photographer — The Oklahoma Daily, University of Oklahoma
Aug 2009 — May 2011
  • Member of the editorial board. Managed a staff of photographers, using time management and decision-making to meet firm deadlines.
Educator — Science Museum Oklahoma
Jun 2006 — May 2011
  • Conducted and developed teaching material for live science presentations. Helped staff overnight events, facilitating science-based projects for children and their parents.