I'm an early-career GenAI & DevOps Engineer based in Chennai, currently pursuing my final-semester B.Tech in Artificial Intelligence & Data Science.
I build production-oriented AI systems — the application, the infrastructure, and the observability that keeps them operable — secure, scalable, and cost-conscious by design.
Infra & platform: Python · Kubernetes · Docker · Terraform · Argo CD/Rollouts · AWS & Azure · Prometheus/Grafana
GenAI: AWS Bedrock · RAG · AI agents · Evaluation · OpenTelemetry
Open to internships in AI, Cloud, DevOps, MLOps, SRE, and Forward Deployed Engineering (FDE).
Selected engineering projects, with measurable outcomes and documented trade-offs.
| Project | What it is | Outcomes |
|---|---|---|
| AutoSRE | Event-driven autonomous SRE agent for Kubernetes — diagnoses from telemetry, tiered remediation with human approval for high-risk actions, verifies recovery | 12/14 incidents resolved · 0 safety violations · ~98% MTTR reduction · $0.02 per incident |
| RAG8s | LLMOps platform for RAG on EKS: streaming inference and batch indexing planes across eight application and connectivity units, with Karpenter-managed Spot compute, GitOps + IaC delivery, OIDC auth with per-user rate limits, and full-stack observability and security | Groundedness: 0.95 · Recall@5: 0.72 · Citation integrity: 0.89 · End-to-end latency: 2.6s · LLM rate limit: 5 req/min per user · 24 alert rules |
| aks-canary-platform | Self-healing canary deployments on AKS (0% → 10% → 100%), where Argo Rollouts automatically rolls back failed k6/Playwright validations | Auto-rollback on failed canary · 0 manual interventions · 0 long-lived secrets · 8 security controls |
| TicketWeave | AI-powered ticket triage for e-commerce support teams on Amazon ECS — deterministic routing, 5-layer security architecture | ~$23/month in staging · sub-5ms inline RAG · 0 critical/high DAST findings · CI 8m → 2m40s via dependency caching |
All reported metrics and results are measured in self-managed development and staging environments, not production. Results are environment-specific. Each repository documents its evaluation methodology, measurement context, setup instructions, design trade-offs, and known limitations.
Always happy to talk about GenAI systems, cloud infrastructure, or opportunities.
- LinkedIn: athithya-sakthivel
- Email: athithya651@gmail.com