Choudhry S.

I build AI infrastructure, DevOps pipelines and GRC platforms that run in production. That covers GPU Kubernetes clusters and self-hosted LLMs, cloud and security for SaaS teams, and compliance products like AMLVault and Manara. BY THE NUMBERS - 30+ GPU Kubernetes clusters built - 30+ AI servers designed and deployed - 60+ LLM deployments - 2 compliance platforms in production (AMLVault and Manara) - Top Rated with 100% Job Success Across university data centres, AI startups, my own company and Upwork clients. WHAT I BUILD 1. AI infrastructure and self-hosted LLMs - GPU Kubernetes on bare metal or cloud: NVIDIA GPU Operator, K3s, Rancher, MetalLB - LLM serving with vLLM, SGLang, Triton and Ollama, running Qwen, Llama and DeepSeek - On-prem GPU servers: hardware selection, Proxmox GPU passthrough, CUDA, monitoring - Inference tuning: quantization (FP8, INT8), tensor parallelism, batching Wrote the IaC for GPU cluster orchestration on KAUST Shaheen III, one of the world's top 20 supercomputers. 2. DevOps, cloud and security - Kubernetes on EKS, GKE, AKS or bare metal, with Helm and ArgoCD - Terraform and Ansible, so everything can be rebuilt from code - CI/CD with GitHub Actions, GitLab CI and Cloud Build - Monitoring with Prometheus and Grafana - Security hardening and technical audits: IAM, secrets, SSL, backups Cut one client's monthly cloud bill from $10K to $4K. 3. GRC and compliance platforms - AMLVault: AML/CTF compliance SaaS for Australian businesses coming under AUSTRAC's Tranche 2 reforms - KYC and KYB through Sumsub, behind a vendor adapter so providers can be swapped - Seven-year, audit-immutable records in PostgreSQL - Manara: production pipeline, full-stack work and the vendor risk module for an enterprise GRC platform - Vendor risk scoring that turns automated security scans into one comparable score Both platforms are in my portfolio below. I also build RAG pipelines, document intelligence with Docling, and LangGraph agents on top of this infrastructure. WHAT CLIENTS SAY "Single handedly built our Kubernetes infrastructure for our GPU setup" "Goes above and beyond and delivers in tight deadlines" "He conducted a thorough security audit, identified critical vulnerabilities, and clearly explained every issue" HOW I WORK - I look at what you already have before suggesting anything new - Everything is written as code, so you can run it without me - You get documentation and a handover call at the end - If something won't work, I tell you early, not at the deadline I'm CTO at Peregrine Ventures AI, and I run my own dual RTX Pro 6000 Blackwell server on Proxmox with GPU passthrough. Tell me what you're building and where it's stuck. I'll tell you what I'd do first.