Backend · Infra · SRE

Megh
Vyas

I build the systems other software runs on. Deployment platforms, real-time AI infrastructure, and the reliability layer underneath.

See the work

Backend & infrastructure engineer

I work a layer below the product. Deployment platforms, real-time voice pipelines, the monitoring that pages you at 3am. The parts nobody sees until they break, which is exactly why they have to be good.

I co-founded Shriem Labs to build retail-tech products and Aexiz Solutions to deliver client infrastructure, started a cloud-computing community, and did the internships in between. Backend, infrastructure, and the AI plumbing that holds it all together.

Where I’ve been

2025 - nowCo-FounderShriem LabsRetail-tech SaaS: allocation engine, multi-tenant schema design.
2026 - nowFounder & Community LeadC3, Cloud Computing CommunityA home for people who run things in production.
2025 - 26Co-Founder & DevOps EngineerAexiz Solutions LLPClient cloud architecture and CI/CD across AWS, GCP, Azure.
2025 - 26SDE-L1, Backend & DevOpsAigenthixAI Interviewer platform: backend, RBAC, LLM pipelines.
2025Salesforce Developer/Admin InternCodiot TechnologiesApex, LWC, and Salesforce Flow automation.
2023 - 27B.Tech, Artificial IntelligenceVijaybhoomi UniversityBachelor's in Artificial Intelligence.

Infrastructure is judged by what does not happen. Here is the order I build in so that nothing does.

Model the failure

Design from how it breaks, not how it demos.

Before the happy path, I map the ways a system falls over: the dropped connection, the slow query, the queue that backs up. The architecture is whatever survives those, not whatever looks clean in a diagram.

Make it observable

If you cannot see it in production, you do not really run it.

Metrics, traces, and logs go in from the first commit, not after the first incident. A system I cannot watch is one I am only hoping works, and hope is not an operational strategy.

Make it boring

The best infrastructure is the kind nobody notices.

The final pass is subtraction: fewer moving parts, saner defaults, automated recovery. Excitement in an on-call channel is a bug. The goal is a system so quiet you forget it is load-bearing.

Six of them. Hover to look.