Muhammad Ammar — DevOps & Linux System Administrator
I build and run Linux infrastructure that has to stay up, and I host AI agents and automation on top of it. Virtualization, storage, cloud, containers, web hosting and self-hosted AI systems, owned end to end from the bare metal to the running app.

AlmaLinux, CentOS, Ubuntu and Debian. Provisioning, hardening, kernel tuning, patching and keeping servers healthy.
GitHub Actions, Jenkins and Drone. Automated build, test and deploy pipelines with zero-downtime releases.
Containerized apps behind NGINX with SSL. 30+ microservices across 10 to 15 servers, Compose and registries.
Production clusters with rolling updates and rollbacks, health checks, and a repeatable deploy pattern across services.
AWS and GCP with Terraform, plus VMware vSphere, ESXi and vSAN. Infrastructure as repeatable code.
Server hardening, fail2ban, firewalls, Wazuh log monitoring, VPN and SPF/DKIM/DMARC email security.
2024 to present · Islamabad
2022 to 2024 · Lahore
2021 · Remote · Cyprus-based company
The infrastructure code, hardening playbooks and ops scripts behind this work, with tfsec, checkov and shellcheck running in CI, are open to read on GitHub.
Read-only audit of two AWS accounts across 14 regions. Half the S3 bill turned out to be request charges, not storage: a sync job was firing about 370,000 list calls a day to copy 7 new files. Fixed for roughly $660 a year. The same pass found 7 buckets open to the internet across 2.6 TB and a production database reachable on its public port.
A vCenter cluster of 4 ESXi hosts, 6 disks per host plus an SSD for cache. vSAN over dual 10 GbE gave a 6 TB pool running 30 production VMs, with backup and DR.
Two MinIO clusters, three nodes each, with erasure coding, an NGINX load balancer for a single URL, and site replication between the clusters for high availability.
Rebuilt the full performance stack on a busy WHMCS install. Time to first byte dropped from about 4 seconds to 54 ms, with no disruption to the live site.
A voice platform for more than 300 users on Docker, Asterisk and Redis, with zero-downtime continuous delivery.
Built serverless automation with AWS Lambda, and set up a CloudFront CDN with caching and IAM-scoped access for fast, secure content delivery.
Took a cold domain to verified inbox delivery. DKIM, a custom MAIL FROM subdomain so SPF aligns to the sending domain, DMARC, and a configuration set logging all nine event types to CloudWatch. The sending key is scoped to two actions and locked to three From addresses, so a leak cannot send as anything else.
Automated build and deploy pipelines with GitHub Actions, Jenkins and Drone CI for PHP, Node.js and containerized apps. Multi-stage builds, parallel deploys, zero-downtime releases.
Run a production Next.js and Strapi platform on a Linux VPS, with GitHub Actions building and deploying on every push, PM2 keeping the app up, and NGINX and SSL in front.
Moved a live app's PHP/MySQL backend to new hosting with no app changes. Built in parallel, kept API endpoints identical, cut over at 3 AM with a write-freeze. 12,000+ records, zero loss, email continuity, instant rollback.
I host AI agent systems and automation on servers I set up and secure. The agents do the work, I make sure the server stays solid and safe.
A FastAPI service that answers plain-language questions about your own documents. Ingests PDFs, Word files and Notion pages, embeds locally, answers with sources cited and a confidence score.
An MCP server that lets a brokerage's agents pull live MLS data straight from a Claude chat. Connects Claude to the Spark (FlexMLS) API with tools for search, comps, market stats and more.
A nightly pipeline that generates search queries with Claude, runs web search, scores each company for fit, verifies the decision-maker's email, dedupes, and delivers a clean list every morning. Runs unattended on the client's own server.
A 5-agent system that runs cold outreach and Reddit automation on its own. I set up the VPS, hardened it, added SSL and login, and set the limits that keep it safe.
A 3-agent pipeline that pulls leads, adds detail to each one, and writes a tailored message. It runs unattended on a single VPS.
Set up and configured OpenClaw and Hermes agent platforms on client servers, including headless Linux with browser automation, gateway, SSL and OAuth in place.
Built and deployed AI voice agents on Twilio that answer calls, route them, record and log to a CRM, including a Retell-based phone agent that books appointments.
Built a sync that pulls patient records from eClinicalWorks over its FHIR API and pushes them into GoHighLevel through webhooks, hosted in the region the access requires.
A custom AI assistant on a live company site that answers questions and captures leads to the sales inbox. PHP backend, model provider swappable in one line, keys server-side, rate limited.
A Cloudflare Worker that writes a personalized campaign email to a constituent's local representatives with an LLM, served through an embeddable widget with an admin page.
Self-hosted n8n in Docker with SSL, used to connect services and run automated workflows and webhooks.
SSL and domain expiry monitoring, RAG support triage, AI lead enrichment and scoring, an SEO blog engine and a multi-platform social publisher. Each with retries, routing, dedup and approval steps.
A few of the jobs I have finished for clients, each closed five stars. Top Rated · 100% Job Success.
A client's app was running on a server ten times bigger than it needed. I measured the actual peak load, sized the replacement, then moved the filesystem, database and SSL across to it with no data loss and no app changes. Hosting went from about $96 a month to $7.
Several domains on one OpenLiteSpeed server were serving the wrong SSL certificate. I traced it to SNI and vhost mapping, rebuilt the per-domain bindings, cleared cache. All sites back on the right cert, no downtime.
Set up and tuned a ClickHouse database holding 5.3 billion rows (83 GB) so analytical queries run fast at production scale.
Recovered and fully rebuilt a Magento store after data loss, and brought the site back online.
Deployed a self-hosted Supabase platform with Docker, PostgreSQL and NGINX, with SSL and a clean setup the client could maintain.
Set up SendGrid with a subaccount, domain authentication and DMARC so a client's mail landed in the inbox instead of spam. Delivered the same day.
Pointed a GoDaddy domain to AWS Lightsail, set up a WordPress instance with SSL, and configured professional email accounts.
Moved a trading VPS from one provider to another with the full setup intact and no data loss, running the same on the new host.
Set up a Tailscale network so a client could reach a laptop, NAS and tablet securely from anywhere, no ports opened. Configured an exit node to route traffic through one trusted machine.
Real feedback from Upwork clients, every one of them five stars.
"So fast, perfect deliverable. Delivered exactly what we needed, faster than expected. I would definitely work with Muhammad again. Great job."
"Muhammad is genuinely great at what he does, he was given a brief and executed it perfectly. Always there when I needed to ask a question. I would 100% recommend and will be doing further work with him."
"Muhammad is very zen, which is what you need in the delivery of mission-critical dev-ops analyses and resolves. A solid professional."
"Reasonable price, quick communication, great job. What's not to like? Thank you Muhammad."
"Great working with Ammar. He completed our website hosting and server migration successfully on-time and we got the website live with no issues."
"Exceptional service and knowledge. He has provided ongoing support for months, worked through all my special requests and always delivered great work. Highly recommended!"
My MinIO project was featured in the MinIO newsletter (September 2024): two clusters with three nodes each, erasure coding so data survives a node failure, an NGINX load balancer for a single access URL, and site replication so data stays available even if one cluster goes down.
I treat security as part of building a system, not something added at the end. I am trained as a SOC analyst and keep building on the blue team side.
Deployed Wazuh as a SIEM for central log collection, custom alert rules, agent management and active threat detection across servers. I harden Linux with kernel tuning, fail2ban, firewalls and patching, set up VPN and certificate-based access, and handle email security with SPF, DKIM and DMARC. Hands-on labs in log analysis, malware basics and privilege escalation, mapped to MITRE ATT&CK.
EC-Council Certified SOC Analyst (CSA v2)
EC-Council SOC for Blue Teaming
arcX Cyber Threat Intelligence 101
TCM Security Linux Privilege Escalation
Hack & Fix Certified Phishing Prevention Specialist
Hack & Fix Cybersecurity Career Starter
EC-Council Security Automation with Ansible
EC-Council Open-source Identity & Access Management
cPanel WHM System Administrator II (CWSA-2)