HB
Book Audit

Hasan Butt · SRE and platform engineer

I help teams run reliable cloud and AI systems.

I work on Kubernetes platforms, cloud costs, and model serving. My work includes tracing production failures, choosing capacity, and leaving the team with code and runbooks they can maintain.

Book a free 20-minute callExplore my work

Based in Pakistan. Working hours and response expectations agreed with each team.

Production Systems I Have Built

Real systems I built and deployed. Each one solved a specific engineering problem. Click any project for the full case study.

MLOps & GPU ServingFreelance client

Self-Hosted LLM Inference on Bare-Metal GPUs

Self-hosted Llama 3.1 70B on four A100 GPUs, with quantized workers and a FastAPI gateway.

82%Reduction in listed monthly charges; hardware and labor excluded
llama.cppvLLMNVIDIA A100CUDA 12.2+9
Explore GPU Architecture
Platform & SREFreelance client

EKS at Scale: Karpenter Autoscaling & JVM SRE

Reworked EKS scheduling and JVM memory limits to address slow capacity changes and container restarts.

<82sScale-Up Latency (Down from 4.5 Minutes)
AWS EKSKubernetesKarpenterHelm+8
Inspect Karpenter Blueprint
Platform & SREBuilt at QisstPay (2022–2024)

Hybrid Cloud Infrastructure for a BNPL Fintech

Unified AWS + GCP infrastructure for a fintech handling 1M+ daily transactions across 17 microservices under PCI DSS audit.

1M+Daily Transactions on Infrastructure I Operated
AWS ECSGCP Cloud RunTerraformMySQL+10
DEMOInspect PCI DSS Infra
AI & RAG PlatformsFreelance client

Multi-Stage Legal RAG with Azure AI Search

Built a 5-layer retrieval pipeline over 3,145 dense legal documents, fixing real failure modes at each step.

0.98Ragas Faithfulness Score (Up from 0.72)
PythonFastAPINext.jsAzure AI Search+8
Review RAG Pipeline
Platform & SREFinTech Client Engagement

Payment API Observability & Latency SLOs

Consolidated fragmented metrics into Datadog APM with strict latency SLOs, cutting recovery time from 40+ min to under 9 min.

78%MTTR Reduction (99.99% Peak Availability)
DatadogAWS LambdaPythonDocker+4
Inspect Architecture
View all 5 projects
Interactive FinOps ROI EstimatorIllustrative scenario

Cloud & GPU Infrastructure Savings Calculator

Explore an illustrative savings scenario. These assumptions are not a quote or an analysis of your actual bill.

Select Primary Cloud Platform:
Monthly Cloud & AI Spend$35,000 / mo
$5,000
$100,000+
Dedicated GPU Fleet / Compute Nodes4 GPUs / Nodes
0 (CPU only)
16+ GPUs
Current Architecture & BottleneckAWS Profile
Illustrative savingsFormula: 45% of your bill × 65% assumed reduction, plus $1,800 per GPU/node. Capped at your bill. Migration and operating costs excluded.
~50% Cost Cut
Est. Monthly Reduction$17,438Before migration and operating costs
Illustrative annual savings$209,256Direct bottom-line savings
Payback on $2,500 Audit~0.6 weeks
Reliability targetAgreed after review
Paid audit guarantee$2,500 audit: $10K annual waste in 14 days or refund
Direct Calendar BookingZero sales pitch · Engineer-to-engineer
Remote Reliability

How I Work with US, Canada & Australia Engineering Teams

Here is how I organize project work, communicate decisions, and hand over the system to your team.

Working hours agreed together

We agree a working window, scheduled check-ins, and response expectations before starting. Live debugging and asynchronous updates both have a place.

Scheduled overlap in your timezone
Written progress updates
Escalation coverage defined in scope

Written decisions and tradeoffs

I document the options, costs, and failure modes behind a proposal. That gives your team something concrete to review before implementation.

Detailed Architecture Decision Records (ADRs)
Written Tradeoff & Failure Analysis
Loom Video & Async Walkthroughs

Access and security controls

I work with your team on scoped access, short-lived credentials, secrets management, encryption, and audit logging. The required controls depend on your data and assessment scope.

HashiCorp Vault & AWS Secrets Manager
Private VPC Subnets & Bastion Zero-Trust
Audit Logs & Immutability

Code and runbooks your team owns

Everything is codified in Terraform / Kubernetes manifests with complete Mermaid diagrams, structured READMEs, Grafana telemetry dashboards, and interactive runbooks. Your internal team owns 100% of the repository and IP.

100% Infrastructure as Code (IaC)
Mermaid Architecture Diagrams
Step-by-Step Incident Runbooks

A clear working window and escalation path

Working remotely from Pakistan, with overlap agreed for each engagement.

Book Free 20-Min Infra Audit
Hands-On Tooling

Interactive Architecture Labs & Simulators

Hands-on technical artifacts replicating real production challenges: GPU serving bottlenecks, FinOps calculators, and Kubernetes self-healing.

bin/lab --id=incident-response

Kubernetes Incident Response Simulator

Automated detection, triage, and recovery in under 42 seconds

The Problem

The production payment pod hit an out-of-memory crash in a 12-node EKS cluster. Liveness probes failed, Horizontal Pod Autoscaler could not scale because the node pool was at capacity, and over two hundred non-critical alerts created notification fatigue. The on-call engineer had to manually SSH into nodes, identify the failing pod, restart it, and verify recovery. That manual cycle averaged forty-five minutes per incident.

Recovery Time
42sfrom 45m
KubernetesPrometheusGrafana
LAUNCH LAB →
bin/lab --id=finops-optimizer

AWS FinOps Cost Optimizer

Line-by-line cloud cost audit that cut $40,920/year

The Problem

The client AWS bill was $6,400 per month for a mid-sized SaaS platform. Contributing factors included over-provisioned EC2 instances in ap-southeast-1 instead of us-east-1, three NAT Gateways routing S3 and DynamoDB traffic across public endpoints at $0.045/GB, absence of a CDN forcing all API responses through the ALB, an idle db.r5.xlarge RDS instance running overnight, and orphaned storage (three unattached 500GB EBS volumes and two unused Elastic IPs).

Annual Savings
$40,92053% cut
AWSCloudWatchCost Explorer
LAUNCH LAB →
bin/lab --id=cold-start-optimizer

18GB ML Model Cold Start Optimizer

Reduced serverless cold start from 2+ minutes to 8 seconds

The Problem

An 18GB model container on Google Cloud Run required over two minutes to cold start because it downloaded model weights from Google Cloud Storage on every scale-from-zero event. Peak memory hit 28GB against a 32GB container limit, causing 504 Gateway Timeouts for initial requests after idle intervals.

Cold Start
8sfrom 2m+
GCP Cloud RungcsfusePython
LAUNCH LAB →
bin/lab --id=terminal-emulator

Infrastructure CLI Emulator

Web-based terminal simulating kubectl, docker, and terraform commands

The Problem

Onboarding engineers on production clusters created operational risks and configuration drift. Provisioning isolated cloud sandboxes cost $800 monthly per engineer and required ongoing maintenance.

Onboarding
3 Daysfrom 2 weeks
ReactTypeScriptNext.js
LAUNCH LAB →
Explore All Interactive Labs & 3D Topology →

Testimonials

What clients say after working with me on system infrastructure and MLOps pipelines.

Upwork
“Hasan delivered a solid end-to-end build — from scraping to deployment. Communication was excellent throughout, and he handled scope changes gracefully. Would hire again without hesitation.”
Mark T.Product Owner, UK Planning PortalData Automation & Supabase Deployment
Upwork
“Hasan did an excellent job installing and configuring PostgreSQL and K3s on our bare-metal servers. His Ansible automation was clean and well-documented. He also coached our team on DevSecOps best practices.”
James R.Systems Administrator, Enterprise ClientDevSecOps & K3s Deployment
Upwork
“Hassan quickly deployed our scikit-learn model on SageMaker with a clean endpoint setup. Very responsive and delivered ahead of schedule. The documentation he provided was thorough.”
Chen W.ML Engineer, Tech StartupSageMaker ML Deployment
LinkedIn
“I hired Hasan for the deployment of our AWS infrastructure and he exceeded expectations. His Terraform modules were clean, well-documented, and production-ready. Highly recommend for any cloud deployment work.”
Haseeb K.Business Development Manager, Partner CompanyAWS Cloud Deployment
Upwork
“Hasan was an absolute pleasure to work with. He quickly understood our requirements for the legal research platform and delivered a clean, production-ready backend with LLM integration. His DevOps background made deployment seamless.”
Sarah L.CTO, AI.ioAI Backend Development
Upwork
“Hasan was very friendly, clear, and knowledgeable about RAG architectures. He set up our entire pipeline with pgvector and LangChain, achieving 95% retrieval accuracy on our domain-specific dataset.”
David M.Founder, AI StartupRAG Pipeline Development
Upwork
“Hasan did an outstanding job helping me with building our CI/CD pipeline from scratch. He set up GitHub Actions, Docker builds, and deployment automation. Our release cycle went from weekly to daily.”
Ahmed K.Engineering Lead, SaaS PlatformCI/CD Pipeline Setup
View all reviews on Upwork
100% Job SuccessTop Rated18 Projects
Hasan Butt — Senior MLOps, SRE & FinOps Consultant

Let's Fix Your Cloud Spend

● Available for Contract · US/CA/AU Hours

Whether you need to eliminate runaway GPU compute bills, build a reliable multi-stage RAG pipeline, or stabilize Kubernetes operations, I am ready to help.

contact@hasanbutt.com
Pakistan · Remote · PST/EST/AEDT Overlap
Direct Scheduling:Book Free 20-Min Infra Audit

Guaranteed: No sales pitch · Actionable findings in 20 minutes

Send an Async Inquiry

Prefer email? Describe the problem and your current stack. This form prepares a draft in your email app.

NDA & IP Protection standardResponse windows agreed per engagement