Hi, I'm Kevin Pham

Hardware engineering lead designing Block's racks of compute. Currently writing a series for GPU cluster infrastructure and electrical design for datacenters.

A brief timeline

  1. 2022 – Present

    Staff Electrical Engineer at Block (parent of Square): Technical lead for all things compute: design for hyperscale air and liquid-cooled data centers, modular prefab compute, MV/LV power systems, design for LVRT and AI-driven LCL oscillations. 100+ MW deployed with a team across US, Canada, and APAC.

  2. 2026 – Present

    Senior Advisor, AI Infrastructure at micro1: Reviews and sets up ECAD, MCAD and robotics AI training pipelines for frontier AI labs.

  3. 2019 – 2022

    Lead Hardware Engineer at Microsoft: Led electrical design for Project IVAS, a US Army augmented reality headset, including the MLB, custom high-speed interconnect and its test.

  4. 2015 – 2019

    Senior System Integrator at Apple: Senior electrical engineer for system, capacitive-sense and optical design on AirPods (over 50M units) and HomePod (over 1M units).

  5. 2016 (part-time)

    Masters in Electrical Engineering and Machine Learning at Stanford University: Completed part-time while working at Apple and Microsoft.

  6. 2012

    Bachelors in Electrical and Computer Engineering at The University of Texas at Austin: Included 18 hours of graduate coursework in computer architecture and VLSI design.

Latest

All 56 posts
Ctrl K

Series

Datacenter electrical design 6.1 Rack power plan 6.1 Power plan 6.2 Energy sources and siting 6.2 Sources 6.3 Interconnection and substation 6.3 Interconnect 6.4 Redundancy architectures 6.4 Redundancy 6.5 Short-circuit and coordination 6.5 Short-circuit 6.6 Busway and 415 V 6.6 Busway 6.7 UPS and batteries 6.7 UPS 6.8 Grounding and power quality 6.8 Grounding 6.9 GPU load transients 6.9 GPU load 6.10 On-site generation 6.10 On-site gen 6.11 Generator and battery response 6.11 Gen + BESS 6.12 Microgrids and black start 6.12 Microgrid 6.13 Power-system studies 6.13 Studies 6.14 Vendor terms and bridging power 6.14 Vendors 6.15 Metering and PUE 6.15 Metering 6.16 Commissioning documents 6.16 Commissioning 6.17 Arc-flash basics 6.17 Arc flash 6.18 400 V and 800 V DC 6.18 DC power Part 0 Foundations 0 Foundations Part 1 Slurm 1 Slurm Part 2 Kubernetes 2 Kubernetes Part 3 Industry use 3 Industry use Part 4 ML platform 4 ML platform Part 5 Platform layer 5 Platformlayer Part 6 Bare metal 6 Bare metal Part 7 Cloud + GPU 7 Cloud + GPU Part 8 References 8 References Homelab series Advanced Kubernetes track 2.1 HA and etcd 2.1 Control plane 2.2 Cilium and network policy 2.2 Networking 2.3 RBAC, quotas and isolation 2.3 Access andtenancy 2.4 cert-manager and rotation 2.4 Certificatesand secrets 2.5 A Go webhook and controller 2.5 Admission 2.6 kubeadm and Cluster API 2.6 Clusterlifecycle

The homelab series starts with the basics: Slurm, Kubernetes, a platform layer, bare metal and GPU scheduling. Following it from Part 0 to Part 8 gives a surface understanding of hosting and running GPU servers. After the Kubernetes post, I branch into more advanced topics in a second series, the Advanced Kubernetes track: the control plane, networking, access, certificates and cluster lifecycle. For more advanced topics in datacenter electrical design, I branch after Part 6, at the rack power plan (6.1), and zoom out to the full chain, from the generators and the grid connection to the rack. Feel free to use the diagram above to navigate the posts. Every dot above is a link.