Principal Cloud & AI Platform ArchitectAWS Pro CertifiedAWS Community Builder

I care about the part that fails at 2 AM.

I am Rahul Ladumor, a Principal Cloud & AI Platform Architect from Surat, India. I work with teams that already have real users, real AWS bills, and real production pressure.

Right now that means an enterprise engagement under NDA: standardizing how the client's teams provision AWS, and designing the review gates that let them use AI coding tools on production code.

The work starts where a demo stops being enough. I want to know what fails first, who gets paged, what rollback looks like, and what the system costs when traffic doubles, then fix the answer that worries the team most.

Request the reliability audit Contact form

Problems I owned

Three systems, what I changed, and what it did.

Each links to the full case study with the trade-offs and the before and after.

Enterprise client (NDA), 2025 to now

Enterprise AWS provisioning

Problem: Fourteen VPCs built by hand, each project with its own security posture. A new environment took 2-3 weeks and the quarterly audit returned 47 findings.
My part: With one other senior architect, I designed the reference architecture and its Terraform modules, and ran weekly reviews with eight junior engineers.
Result: Environments now come from reviewed modules in about 45 minutes, and audit findings dropped to 3 a quarter.

Read the case study

Enterprise client, 2024

Enterprise knowledge search

Problem: 2M+ internal documents across PDFs, Confluence and Slack. Engineers spent 30+ minutes finding answers that were already written down.
My part: As architect on a team of three engineers, I designed the RAG pipeline on Amazon Bedrock: ingestion, chunking, OpenSearch Serverless retrieval, a semantic cache and source attribution.
Result: Answers in under 10 seconds with their sources shown, and the cache cut repeated LLM calls by about 60%.

Read the case study

Enterprise client (NDA), 2026

AI-assisted delivery

Problem: The client wanted AI coding tools working on real tickets. Without repo standards, application context and review gates, AI writes code that compiles but does not fit the system.
My part: I designed the governed Claude Code workflow: written standards, per-application knowledge bases, commands, Jira context, pull-request review gates and a phased rollout plan.
Result: The design keeps human review as the quality gate: AI drafts code and tests, and a developer reviews each pull request before merge.

Read the case study

Current signal

Where I am right now.

Current roles and credentials. The client work is under NDA, so it is described here rather than named.

What I'm working on now

Senior AWS Solutions Architect at an enterprise consultancy, under NDA

Forward-deployed delivery model: embedded discovery, implementation, integration, rollout, adoption, and production ownership

LLM Infrastructure Trainer on an infrastructure-as-code RLHF program (NDA)

5x AWS Community Builder (Serverless)

IIT Roorkee PG Certificate in Agentic AI, GenAI & ML

Writes up production AWS fixes from real failures

Why I build this way

Production systems need boring answers to hard questions.

Production does not care how good the demo looked.

What fails first matters more than what looks clean in a diagram.

Rollback is part of architecture, not release management paperwork.

A platform should reduce decisions for product teams.

GenAI quality problems usually start before the model call.

Career proof

Full timeline across cloud, backend, DevOps, platform, and AI infrastructure.

Every role since 2016, newest first. Some ran in parallel as contracts or part-time work. Where the work ran through a firm or agency, the row says so; where an NDA applies, the client is described instead of named.

Enterprise consultancy, under NDA

Senior AWS Solutions Architect

2025 - Present

Forward-deployed with enterprise stakeholders to turn ambiguous cloud needs into end-to-end AWS architectures. Built reusable IaC with Terraform and CDK plus multi-account guardrails, cutting environment setup time by 70%. Reduced cloud spend about 20% through analytics-driven right-sizing.

AI training program via a delivery partner (NDA)

LLM Infrastructure Trainer

2025 - 2026

Validated AI-generated AWS CDK and CloudFormation for security and cost on an infrastructure-as-code RLHF program. Built reusable TypeScript CDK constructs aligned with Well-Architected principles, cutting standardization time about 30%. Provided RLHF feedback that raised the quality of generated infrastructure code.

NTT (Tokyo)

Senior Software Engineer (Contract)

2025

Short AWS platform contract in Tokyo. Supported NTT's cloud infrastructure initiatives in a customer-facing engineering capacity, contributing to AWS-based platform development for enterprise cloud capabilities.

topmate.io

Cloud & AI Career Coach (Part-time)

2024 - 2026

1:1 mentoring for cloud professionals and engineers. Specialized in AWS Architecture, Cost Optimization, and guiding engineers transitioning to AI/ML and GenAI career paths.

Freelance

Independent Cloud & AI Consultant

2023 - 2026

Forward-deployed into client problems: discovery, architecture review, implementation, and post-launch guidance. Designed serverless architectures with selected 30-40% operational cost reductions. Integrated LLM-powered features using AWS Bedrock for fintech and e-commerce platforms.

ProdigyBuild

Lead Platform Engineer

2023 - 2025

Worked close to product and engineering users to architect a serverless SaaS platform using Python, Lambda, API Gateway, and DynamoDB, with a 99.9% availability target and 40% cost reduction. Integrated AI agents for automated code generation.

ProtectOnce

SDE-II

2022 - 2024

Built developer-first API security on AWS serverless. Developed Python backend services for automated threat detection and eBPF kernel-level agents for runtime protection. Shipped a React security dashboard with real-time threat visualization and improved deployment efficiency by 25% through CI/CD gates.

QSR e-commerce platform + Qraved (via Imaginato)

Senior Software Engineer

2022 - 2023

Serverless engineering for two Imaginato clients. For a major QSR e-commerce platform in Thailand, migrated a PHP monolith to event-driven Node.js, Go, and Kafka services processing 5M+ orders/month at under 200ms latency. For Qraved Indonesia, rebuilt the discovery and loyalty platform on Lambda, DynamoDB, and GraphQL, tripling throughput, cutting infra cost about 30%, and lifting CTR 18% and DAU 22% with Amazon Personalize.

PrimeLab

Blockchain Developer

2022

Short blockchain engagement building backend services in Go with DynamoDB. Implemented on-chain data workflows and API integrations for a decentralized platform.

NDS Global

Senior Software Developer

2021 - 2022

Built conversational AI solutions for enterprise clients using AWS Lex and Azure Bot Services. Designed NLU models and integrated chatbots with enterprise CRM and ticketing systems.

AppGambit

Full Stack Cloud Developer

2020 - 2021

Worked in AppGambit's custom software and AWS consulting environment, operating cloud-native microservices handling 1M+ requests/minute on ECS, EKS, and Fargate. Built real-time audit logging with Lambda, S3, and ElasticSearch.

Bynebits

Software Developer

2017 - 2019

Built scalable web applications using Node.js and AWS serverless. Developed investment analytics platform. First exposure to serverless architecture and AWS Lambda.

Creative Infotech

Software Engineer

2016 - 2017

First professional role. Built and maintained PHP web applications and MySQL backends, learning production fundamentals: deployments, debugging, and shipping features for real users.

What I've built

A product, not just engagements.

Case studies describe work done for someone else. This is something you can go and run yourself, today, against your own account.

IdleLens

A read-only AWS waste scanner. It finds idle and unused resources with CloudWatch evidence and a confidence label, and never modifies the account.

Scanners: 19

Access: Read-only IAM, by policy

Setup: CloudFormation role, External ID

First scan: Free, no card

What I work with

The stack behind the production work.

Cloud Architecture & Cost

AWS Cost Audits (30-40% typical, up to 55%)Serverless Architecture (Lambda, API GW, DynamoDB)Multi-Account AWS Organizations & Landing ZonesContainer Orchestration (ECS, EKS, Fargate)Event-Driven Systems (EventBridge, SQS, SNS, Kafka)Multi-Region / Multi-AZ for 99.99% UptimeFinOps: Savings Plans, Reserved Capacity, Right-Sizing

Production AI Infrastructure

AWS Bedrock (Claude, Titan, Llama): production-readyRAG Pipelines: chunking, retrieval quality, latency budgetsAgentic AI with real guardrails and cost controlsLLM Token Optimization and Semantic CachingVector Search (OpenSearch Serverless)AI system observability and failure mode handlingModel evaluation and tracing (RAGAS, LangSmith)MLOps pipelines (SageMaker Pipelines, MLflow, DVC)Realtime voice AI (LiveKit, OpenAI Realtime API)

Infrastructure as Code

Terraform (multi-environment, remote state, modules)AWS CDK (TypeScript): one Stack per service boundaryCI/CD: GitHub Actions, CodePipeline, blue-green deploysMonitoring: CloudWatch, Datadog, X-Ray tracingSecurity: least-privilege IAM, KMS, WAF, GuardDuty

Backend Systems at Scale

Event-driven microservices that survive traffic spikesDynamoDB single-table design for predictable latencyMonolith to serverless migrations (Strangler Fig)API design: REST, WebSocket, streaming responsesPython, TypeScript, Go: production, not prototypesAPI services in FastAPI and FlaskeBPF kernel-level agents for runtime protectionKafka event streaming and PostgreSQL at scale

Community and writing