serviceB2B contractremote · GMT+3
AWS consulting & cloud architecture
Architecture, scale, cost, AI, and data on AWS — from a certified Solutions Architect who's been on the pager for the systems he designs.
What I do
- Design AWS architectures end-to-end — compute, networking, data, security — picked for the actual workload, not the default reference pattern.
- Run architecture reviews against the AWS Well-Architected Framework with concrete, ranked remediation plans.
- Build and rescue scale-critical surfaces: API Gateway, Lambda, ElastiCache, DynamoDB, ALB / NLB, EventBridge — including rate-limiting and throttling tiers.
- Integrate generative AI into production AWS workloads using Amazon Bedrock (Claude, Llama, Titan), Knowledge Bases, Guardrails, and Agents.
- Build RAG pipelines on AWS — Bedrock Knowledge Bases, OpenSearch Serverless / Aurora pgvector / Kendra for retrieval, embedding-model selection, chunking strategy, and eval harnesses that catch regressions before they ship.
- Train, fine-tune, and host models on SageMaker — bring-your-own-container, distributed training, async / serverless inference, shadow deployments. Tune inference cost with Bedrock cross-region inference, provisioned throughput, prompt caching, and small-model distillation for hot paths.
- Wire real-time data into AI — Kinesis / MSK → Amazon Managed Apache Flink → Bedrock for streaming inference, with feature stores and online/offline parity.
- Design AWS data platforms — S3 + Iceberg lakehouses, Glue / EMR / Athena for batch, Redshift / Redshift Serverless for warehousing, DMS for migrations, Lake Formation for governance.
- Model DynamoDB for high-scale workloads — single-table design, GSIs, DAX caching, on-demand vs provisioned, and zero-downtime data migrations.
- Cost-tune AWS spend — right-sizing, savings plans, storage class moves, network egress reductions. Typical engagements pay for themselves.
- Write Terraform / CDK for the resulting architecture so it's reproducible and reviewable.
- Run k6 / Artillery load tests against staging to validate scale under realistic burst patterns.
- Co-organize the AWS Ankara User Group and Cloud Turkey — I stay current with the AWS surface, not just what I shipped last quarter.
When teams hire me
- A new product needs an AWS architecture from scratch and the team wants a sanity check before committing to it.
- An existing AWS workload is failing under load (API Gateway throttles, Lambda concurrency exhausting, ElastiCache CPU spikes) and the root cause isn't obvious.
- AWS bill is growing faster than the product and engineering can't explain why.
- A team wants to add a Bedrock-based AI feature and needs an architect who's done it without lock-in regrets.
- A team is building a RAG / AI agent feature and needs help picking the right vector store, embedding model, chunking, and eval strategy.
- A data platform on AWS is straining — Athena query times exploding, Glue jobs failing, Redshift WLM mis-tuned, DynamoDB hot partitions — and the team needs an outside read.
- A migration to AWS is mid-flight and the cutover plan is wobbly.
- A serverless app needs to survive a marketing push without a 24-hour war room.
Engagement formats
Architecture review — 1 week
Read the IaC, trace request paths, model the load and cost. Deliverable: written Well-Architected-style assessment with ranked risks and concrete fixes.
Build engagement — 3-8 weeks
Design, implement, load-test, and document a specific surface (rate-limiting tier, Bedrock integration, event pipeline, multi-tenant boundary). Pair-programming with your team so they own it.
Cost & scale tuning — fixed-fee
Reduce AWS spend or right-size for next-quarter growth. Scoped against measured savings or a specific load profile.
Incident-response engineering — on retainer
Be the second pager when scale incidents happen. Monthly retainer, 24h response.
Certifications
AWS Certified Solutions Architect — Associate
AWS Certified AI Practitioner
AWS Certified Cloud Practitioner