I design and operate production-grade AWS platforms with automation as the default. My work focuses on performance, cost efficiency, and secure delivery through scale-to-zero patterns, OIDC-based authentication, and modular Terraform architecture.

My focus:

• AWS infrastructure as code (Terraform, remote state, policy-as-code)
• CI/CD pipelines with GitHub Actions and OIDC (no long-lived AWS keys)
• Scale-to-zero patterns (wake on demand, auto-sleep when idle)
• Observability-first systems (metrics, dashboards, saturation visibility)
• Infrastructure platforms built for real-world behavior under load

• Cloud Platforms: AWS (EC2, ECS Fargate, Lambda, API Gateway, RDS, DynamoDB, S3, CloudFront, Route 53, SageMaker)

• Infrastructure as Code: Terraform (remote state & locking, modular design, policy-as-code), tflint, tfsec, checkov

• CI/CD & Security: GitHub Actions, OIDC (keyless authentication), automated plan/apply workflows, least-privilege IAM

• Containers & Orchestration: Docker, Kubernetes (k3s), Helm, ECS Fargate

• AI & GPU Platforms: vLLM serving, concurrency control, bounded queueing, GPU telemetry (DCGM)

• Observability & Reliability: Prometheus, Grafana, CloudWatch, metrics-driven tuning, golden signals

• Architecture Patterns: Scale-to-zero, event-driven systems, cost-aware design, automation-first infrastructure

Repo: https://github.com/rusets/aws-multi-tier-infra

A full 3-tier production-style web architecture on AWS featuring:

• VPC, ALB, and EC2 application tier
• RDS MySQL in private subnets (no public exposure)
• S3 + CloudFront for static asset delivery
• Wake-on-demand orchestration (Lambda → GitHub Actions → Terraform Apply)
• Idle reaper automation that tears down infrastructure when unused
• Remote state backend (S3 + DynamoDB lock table)
• GitHub Actions OIDC for keyless CI/CD

Purpose: showcase a secure, automated, and cost-aware AWS architecture with real infrastructure lifecycle management.

Repo: https://github.com/rusets/gpu-llm-inference-service

Single-node GPU inference platform built around:

• vLLM for high-performance LLM serving on a single GPU
• Custom FastAPI gateway with explicit concurrency limits and bounded queueing
• Deterministic backpressure (429 / 503) instead of uncontrolled latency growth
• Prometheus + Grafana for full observability (latency, queue depth, saturation, GPU metrics)
• Docker Compose stack for reproducible local deployment

Purpose: demonstrate how to operate a GPU-backed LLM service with controlled concurrency, measurable saturation, and production-style observability rather than just exposing a model endpoint.

Repo: https://github.com/rusets/helmkube-autowake-cicd

A compact k3s cluster on EC2, fully automated through Terraform and GitHub Actions:

• k3s node bootstrap via user data + SSM
• Helm-driven application deployment from CI/CD
• Prometheus + Grafana monitoring stack deployed automatically
• Wake/sleep automation to eliminate idle costs
• OIDC authentication (no AWS keys)
• Remote state, IAM roles, monitoring, and bootstrap scripts

Purpose: implement a fully automated Kubernetes environment on AWS with GitHub → Terraform → Helm delivery pipeline and integrated observability.

Repo: https://github.com/rusets/docker-ecs-deployment

ECS Fargate–based container platform with:

• GitHub Actions CI/CD (build → push to ECR → deploy to ECS)
• On-demand provisioning via Lambda + EventBridge
• Terraform-managed VPC, security groups, IAM roles, task definitions, and services
• Remote state backend + OIDC (no long-lived AWS credentials)

Purpose: implement a cost-aware, event-driven container platform on AWS and illustrate when ECS Fargate is the pragmatic alternative to Kubernetes.

Repo: https://github.com/rusets/ml-sagemaker-serverless

End-to-end ML inference pipeline on AWS using:

• SageMaker Serverless for fully managed model hosting
• Lambda + API Gateway for a clean HTTP prediction API
• Static UI on S3 + CloudFront (image upload → prediction result)
• Terraform for complete infrastructure provisioning (IAM, API, buckets, hosting)
• Keyless CI/CD pipeline (GitHub Actions → AWS OIDC)

Purpose: implement a fully serverless, production-style ML inference API with infrastructure-as-code, secure CI/CD, and zero idle compute management.

Repo: https://github.com/rusets/CI-CD-Pipeline-for-Application-Deployment

A CI/CD-focused deployment platform for running a web application on EC2 with:

• GitHub Actions pipelines (build → test → Terraform plan/apply)
• Wake/sleep automation for cost-optimized EC2 usage
• CloudWatch dashboards and alarms for runtime visibility
• Separate Terraform stack for wake/status Lambdas and API Gateway
• Remote state backend + OIDC authentication (no long-lived AWS keys)

Purpose: implement a clean separation between application delivery and infrastructure management, with cost-aware automation and secure, fully automated deployments.

Repo: https://github.com/rusets/ridebot-infra

A fully serverless backend powering a Telegram-based ride request system, built using:

• API Gateway + Lambda (event-driven HTTP backend)
• DynamoDB for state management and request tracking
• Amazon Location Service for geolocation processing
• Terraform-managed IAM, routing, and infrastructure configuration

Purpose: implement an event-driven architecture with chat-platform integration, leveraging a pure pay-per-use serverless model with no idle infrastructure.

Repo: https://github.com/rusets/rusets-portfolio

Infrastructure powering my personal website https://rusets.com, built with:

• Private S3 + CloudFront (Origin Access Control)
• Route 53 + ACM (DNS validation & HTTPS)
• Terraform-managed infrastructure (remote state, modular design)
• GitHub Actions with OIDC for fully keyless CI/CD

Purpose: implement a secure, production-grade static hosting stack with least-privilege access, automated deployments, and zero long-lived credentials.

Repo: https://github.com/rusets/rdservicepros-site

Production static hosting stack for my small business RD Service Pros (Navarre, FL), built with:

• Private S3 + CloudFront distribution
• HTTPS via ACM
• GitHub Actions for automated deployments and cache invalidation
• Cost-efficient, low-maintenance infrastructure

Purpose: implement a secure and automated static hosting architecture suitable for a real small business, with minimal operational overhead and predictable cost.

• AWS Certified Solutions Architect – Associate
• AWS Certified Developer – Associate
• AWS Certified AI Practitioner
• AWS Certified Cloud Practitioner
• Linux Essentials

• Hands-on experience designing and operating production-grade cloud platforms:
  • Terraform remote state backends with locking and state isolation
  • Multi-account and multi-domain AWS architectures
  • IAM least-privilege design and continuous security hardening
  • Observability-driven performance tuning and capacity planning
• Background in hardware and high-performance GPU operations prior to transitioning fully into cloud and platform engineering

• Portfolio: https://rusets.com
• LinkedIn: https://linkedin.com/in/rdashkin
• Credly: https://www.credly.com/users/rdashkin/badges