U40–U41salman.local
I lead the infrastructure behind large-scale cancer research.
Hi, I'm Salman. I lead platform engineering at the University of Chicago's Center for Translational Data Science: Linux, Kubernetes, GPU clusters, and the people who need them to work on a Monday morning.
All systems operational
Location: Chicago
I run the production platform and the engineering team behind it, and I spend most of my time on reliability, incident response, and the architecture decisions that keep both working. The details are in the work section below.
U34–U37Work · five bays
Building platforms. Leading the teams behind them.
2020 – now
Principal Platform Engineer
University of Chicago · Center for Translational Data Science · Chicago, remote
I lead platform engineering for the NCI Genomic Data Commons, a cancer-genomics platform providing researchers worldwide with access to open and controlled-access data. I own platform architecture, releases, and incident response, and lead the engineering team day to day. My responsibilities also include a Slurm cluster that bursts to AWS Spot, CI/CD across hundreds of repositories, observability, and enterprise vendor contracts.
One project I'm especially proud of is rebuilding our GPU infrastructure. We moved from manually allocated V100 and H100 GPUs to a shared, zero-trust, FIPS-hardened Kubernetes AI platform with a self-hosted model gateway. I defined the architecture and reviewed the designs; the team built the platform. We then made self-hosted coding assistants available to every engineer, with inference and code kept within the center's infrastructure.
2018 – 2020
Lead Pre-sales & Systems Engineer
Motorola Solutions · Muscat, Oman
Lead pre-sales engineer on a $250M+ public-safety program for a key Middle East customer, and business development across the Gulf.
Separately, I redesigned a multi-million-dollar national police operations-centre proposal, replacing the default end-to-end IBM stack with an open-source architecture at roughly a fifth of the cost. My proposed architecture was selected over the IBM alternative. The project finished $4M under budget, and I led the nationwide rollout: a 400-VM server farm, petabyte-scale storage, nationwide IP networks over MPLS and microwave, and the 9999 emergency telephony system.
2017 – 2018
DevOps Engineer
EasyTaxi · Lahore, Pakistan
Dev, test and prod on AWS with Terraform and Puppet; Jenkins pipelines on Spot fleets; New Relic and Datadog. Led the engineering side of the Orange Telecom collaboration for the platform rebrand.
2017
Solutions Engineer
NETS International · Dubai, UAE
Built the company's internal Red Hat OpenStack cloud; designed SDN and SD-WAN proofs of concept; consulted on a 62-site NFV rollout for a major UAE telecom.
2015 – 2017
Technical Solutions Engineer
PLUMgrid (SDN; acquired by VMware) · Islamabad and Santa Clara
Ten OpenStack and SDN deployments across the US, Japan and Switzerland. Lead developer of the PLUMgrid Sandbox, a nested-virtualization playground on Red Hat OpenStack and AWS. Maintained the OpenStack-Ansible Neutron integration. Trained engineers at NASA and Rackspace in San Antonio.
University of Chicago Booth School of BusinessMBA, Honors.
NUST, IslamabadB.E. Electrical Engineering.
CompTIASubject matter expert for the DevOps+ and AutoOps+ certifications.
U31–U32Lab · things built for their own sake
The lab shelf.
Home lab · agents
A self-hosted agent with explicit permission boundaries
A self-hosted agent that browses the web, runs approved commands, and talks to me over Discord. Network and command permissions are restricted by default. The write-up covers the architecture, tradeoffs, and remaining gaps.
Read the write-up →
Home lab · life OS
A vault with a staff of agents
Scheduled agents turn my calendar, inbox, and Obsidian notes into morning briefings, meeting preparation, and updated task lists. They run on an always-on Mac mini and preserve context between sessions, using self-hosted or frontier models according to the task.
PLUMgrid · sandbox
A whole OpenStack cloud in an afternoon
A full OpenStack-plus-SDN environment running inside nested virtualization on AWS, built so customers and partners could try the product in an afternoon instead of a procurement cycle. Later used to train engineers at NASA and Rackspace.
This site
A server rack, in CSS
No framework, no build step, no tracking. The rack is HTML and CSS; the console is about sixty lines of JavaScript. Fonts are the only thing loaded from anywhere else.
Source on GitHub →
U30Writing
Notes from the lab.
Feb 2026
What you'll learn: how to start an agent from zero trust and open permissions one at a time, where an egress proxy and an approval gate belong, and which risks are still open once all of that is in place.
Next on the shelf: what it actually costs to serve an open-weight model per token on hardware you own, with the depreciation included. More at salsikandar.substack.com.
U29KVM console · tty1
Console.
salman.local login: visitor
Last login: just now from somewhere on the internet
visitor@salman.local:~$ whoami
Salman Sikandar. Platform engineer, Chicago. Type help for commands.
Try ls, nvidia-smi, cat writing, or cd work.