# LAB
patterns running in production.
Each blueprint is an AWS architecture pattern we've designed, built, and run in production — not a whitepaper. When a client problem maps to one of these, we adapt — not rewrite. Least-privilege IAM, KMS-encrypted data paths, zero-downtime deploys, and observability are already baked in.
- AWS · EMR · Spark
EMR platform for mixed teams
Visual pipeline authoring, batch jobs on EMR Classic, interactive Spark on EMR Serverless, an S3 file explorer, and per-team cost reporting — one web app for engineers, data scientists, and analysts.
- EMR
- Spark
- React Flow
- Serverless
→ /lab/emr-orchestration
- AWS · Serverless · IaC
Billions-per-day audience distribution
Fully automated, serverless data distribution at scale. Step Functions orchestrates every step end-to-end, on-demand EMR and Glue handle the data plane with zero idle cost, codified end-to-end as IaC.
- Step Functions
- EMR
- Glue
- Aurora
→ /lab/audience-distribution
- AWS · LangGraph · Agents
Agentic workflows on AWS
Multi-tenant agentic workflows — a two-phase planner/supervisor on LangGraph, schema-validated plans compiled to deterministic StateGraphs, typed tool registry. Encryption, least-privilege IAM, and observability baked in.
- LangGraph
- Lambda
- DynamoDB
- Cognito
→ /lab/agentic-orchestration
- AWS · Private AI · Self-hosted
Private LLM platform on AWS
Companies run open-weight models on dedicated GPU capacity in their own AWS account. Zero standby cost, production-grade reliability, full IaC reproducibility — your models, your data, your VPC.
- Ollama
- GPU EC2
- Lambda
- SQS
→ /lab/llm-inference
want one of these in production?
30-min discovery call. we adapt the blueprint, we don't resell it.
book-call// or write: hello@saloid.com · gräfelfing · de