Mohamed S.

Backend and infrastructure engineer, AWS is home. The thing I keep getting pulled into lately is cost, because most AWS waste hides behind stuff that works and nobody questions the service that runs fine. I cut a production search setup 85%, Aurora at $4,200/mo down to DynamoDB plus OpenSearch around $630. Nobody had touched it in a year because it was "working." Found CloudWatch logs double-shipping at $2,766/mo because two teams set up logging and nobody checked for overlap. Found metric streams burning $1,900/mo sending everything unfiltered across regions. That pattern got annoying enough that I built CostPatrol, an AWS cost tool with 128 detection rules across 38+ AWS services that scans an account in about a minute and hands you the exact fix commands. The first version shipped in two weeks. So if your AWS bill is climbing and nobody can tell you why, that's the work I want. Connect a read-only role, I scan, and you get a ranked list of what's leaking and what it costs. No guessing. The cost instinct comes from running production for years, not reading about it. Some of what I've built: - Built the search behind every listing on a major UK ticketing platform from scratch. Under 50ms with filters across 100,000+ performances, and schema changes with zero downtime. - Flash sales at tens of thousands of requests a minute across 80+ venues. 100x DynamoDB throttling reduction by consolidating scheduled work. - Deploy times from 45 minutes to under 5, across 40+ brands and 4 AWS regions. Feature rollouts from 10 minutes to 5 seconds after moving the frontend to Next.js on App Runner. - AI chatbot on Bedrock and Claude, RAG with PII redaction, automating 80% of inquiries. Claude Code is my primary dev tool now. I write the spec, it does the implementation, and I gate the parts that actually bite, auth, money, security, infra. This year I shipped four production products solo this way: an AWS cost SaaS, an institutional deal-verification platform, an AI nutrition coach, and a quant trading engine. The language stopped being the hard part. What breaks at 3am, what needs a human looking at it, that's the part a model doesn't give you. So the stack is a detail. Node and Python are home, but I'll reach for whatever the job needs. Industries I've shipped production in: fintech (KYC/AML onboarding, payments, multi-tenant auth with Cognito), wealth and investments (unit-trust onboarding, fund operations), healthcare (HIPAA, KMS, VPC isolation), ticketing and entertainment (flash-sale load, multi-region, multi-brand), legal-tech (dispute resolution, OCR, e-signatures), e-commerce, petroleum (Fortune 500 dashboards, SAP, Step Functions ETL), IoT (real-time telemetry, leak detection), and blockchain (Hyperledger identity, audit trails). What I take on: AWS cost audits and FinOps (not only AWS, I've trimmed observability tool spend too), backend and API builds (Node, Python, GraphQL, REST), DevOps and infrastructure-as-code (Terraform, CDK, Serverless, CI/CD, multi-account), observability and alerting as code (Pulumi, New Relic, NRQL), AI and LLM work (RAG, agents, Bedrock, Comprehend, Textract, vector search), GEO and AEO optimization for AI search visibility, and architecture, breaking monoliths into event-driven microservices that hold up under load. What I reach for most: AWS (Lambda, ECS, Fargate, DynamoDB, EventBridge, Step Functions, SQS, CloudFormation, CDK, Terraform), Node.js/TypeScript, Python, Claude/Bedrock, LangChain, FAISS, OpenSearch, PostgreSQL, Aurora, Redis, GraphQL, Stripe, Cognito, Cloudflare Workers, Pulumi, New Relic, React/Next.js. Send over what you're dealing with, the bill that doesn't add up or the system that keeps paging someone, and I'll tell you how I'd approach it.