Avinash Sharma — ML Systems Engineer

Avinash Sharma — ML Systems Engineer AVINASH_SHARMA I build AI systems that survive production. 3 years hardening infrastructure at scale. Now applying that discipline to machine learning. → View My Work → Read My Story Available for ML Systems & AI Infrastructure roles avinash@portfolio ~ 3+ Years Building 5+ Projects Shipped 99.9% Uptime SLA Maintained THE_TRANSITION SRE → ML Systems Engineer Why infrastructure thinking makes better AI engineers PHASE 01 2022 — 2024 ── THE INFRASTRUCTURE YEARS ── - Kubernetes at scale - Multi-cloud (AWS · Azure · GCP · OCI) - Terraform IaC - CI/CD pipelines - Observability stacks - On-call discipline > "If it breaks at 2 AM, you own it." PHASE 02 2024 — Present ── THE AI SYSTEMS ERA ── - ML pipeline observability - Agentic AI orchestration - UEBA & anomaly detection - LLM fine-tuning & RAG - Production model serving - SLA-driven AI systems > "I build AI that doesn't break at 2 AM." Most ML engineers come from research or data science. They know how to train a model. They don't always know how to keep it running. I came from the other direction. Three years of on-call rotations, production incidents, and infrastructure at scale taught me one thing: systems fail in ways you never expect. When I started working on ML systems, I saw the same patterns. Models drifting silently. Pipelines failing without alerts. Inference latency spiking with no visibility into why. My SRE background wasn't a detour. It was the prerequisite. I build AI systems with the same discipline I applied to infrastructure: observable, fault-tolerant, and designed to survive the real world. SKILL_COMPARISON | Skill | Most ML Eng. | Avinash | | --- | --- | --- | | Model Training | ✓ | ✓ | | Production Deployment | ✗ | ✓ | | Observability | ✗ | ✓ | | Fault Tolerance | ✗ | ✓ | | On-call Discipline | ✗ | ✓ | | Infrastructure as Code | ✗ | ✓ | | Multi-cloud | ✗ | ✓ | TECH_STACK Stack Two disciplines. One engineer. ⚡ AI / MACHINE