AWS Advanced Tier Services Partner (Differentiated) · Anthropic Partner · OpenAI Partner
Generative AI
Production AWS engagements VeUP scoped, built, and operated — the proof behind the AWS Generative AI Competency. Public references are named; the remainder are shown as AWS-validated anonymized engagements.
Identity protected
Amazon BedrockAWS Step FunctionsAWS LambdaAmazon Rekognition
Amazon Bedrock
AWS Step Functions
AWS Lambda
Amazon RekognitionGenerative AI Applications · Multimodal Content Moderation
Production Engagement · Anonymized
Global social platform reaches 99.6% moderation accuracy on Amazon Bedrock
54% → 99.58% NSFW accuracy with no fine-tuning · ~$1 per 1,000 images · 11/11 SOW acceptance criteria passed
Read the case study →
Amazon BedrockECS / EKS FargateRDS for MariaDBAWS WAF
ECS / EKS Fargate
RDS for MariaDB
AWS WAFGenerative AI Applications · Agentic NL-to-Data
Sofvie Inc.
Sofvie launches a production AI data agent 28 days early on Amazon Bedrock
Shipped 28 days early · 0 cross-tenant incidents at go-live · MCP tool-use over 800+ APIs with AST-validated SQL
Read the case study →
Amazon BedrockECS FargateRDS PostgreSQLpgvector
Generative AI Applications · Agentic Support (RAG + Live Tool-Use)
Cluster Systems (product: ClusterPOS)
ClusterPOS empowers its support team with an AI agent on Amazon Bedrock
1–2s answers (beat a 5s SLA) · sub-800ms KB retrieval over ~600 docs · managed-services continuation signed
Read the case study →
Amazon EKS (GPU)Snowflake on AWSAWS Step FunctionsAmazon RDS for PostgreSQL
Amazon EKS (GPU)Generative AI Applications · GPU ML Inference Platform
GoBubble
GoBubble migrates 21 workloads from GCP to Amazon EKS with zero downtime
Zero-downtime GCP-to-AWS migration of 21 workloads · production ML inference on dedicated EKS GPU node groups · $381,248/yr projected AWS run-rate (AWS Pricing Calculator)
Read the case study →
Amazon BedrockBedrock AgentsBedrock Knowledge BasesBedrock Guardrails
Generative AI Applications · Guardrail-Hardened Agentic AI
MyCena Security Solutions
MyCena hardens its AI agents with Anthropic Claude on Amazon Bedrock
Claude-on-Bedrock agents with guardrail-enforced prompt-injection protection · benchmarked 3-agent/3-model selection (Haiku latency, Sonnet reasoning) · WAR delivered
Read the case study →
Amazon BedrockAnthropic ClaudeMCP tool serversRAG (optional)
Anthropic ClaudeGenerative AI Applications · Tenant-Isolated CRM Assistant
SalesNOW (Interchange Solutions Inc.)
SalesNOW embeds a tenant-isolated AI assistant with Claude on Amazon Bedrock
Tenant-isolated read-only NL-to-data assistant inside the CRM · MCP tool-calling over existing APIs · Smart Reply suggestions · human-in-the-loop v1
Read the case study →
AWS Well-ArchitectedAmazon AuroraAWS KMS (CMK)Bedrock AgentCore
AWS Well-Architected
Amazon Aurora
AWS KMS (CMK)
Bedrock AgentCoreGenerative AI Applications · SaaS & Cloud Operations
Allstacks
Allstacks hardens its DevOps SaaS with a 47-finding six-pillar WAR
6 of 6 pillars assessed across 57 best-practice questions · 47 prioritized findings (31 High, 16 Medium) · per-pillar risk profile + top-5 strategic remediation roadmap
Read the case study →
Identity protected
Amazon KinesisECS FargateAmazon AuroraAWS Verified Permissions
Amazon KinesisGenerative AI · Governed AI Data Platform
Production Engagement · Anonymized
A governed traveler-memory platform, ready for its first enterprise deployment
Enterprise partners can publish traveler data and retrieve governed profiles through a defined MCP interface, with request provenance and consent state enforced at the infrastructure layer · an audit vault capturing every ingestion, compilation and rendering event as Parquet in Amazon S3, queryable via Amazon Athena · AI prompts and recommendations recorded as first-class governed events, each attributed to the model that produced it and bound to a lawful basis · KMS crypto-shredding so an erasure request reaches AI-compiled data while the immutable audit record survives · the platform demonstrable to enterprise travel partners as a working, auditable, governed-memory system
Read the case study →
Amazon BedrockOpenTelemetryGitHub ActionsBudget guardrails
Generative AI · LLMOps & Model Evaluation
Zenhub
Zenhub gates every model swap on 12,000 real prompts, moving to Amazon Bedrock
Every model call metered, tagged, and budget-enforced through one gateway with fallbacks — deployed and validated in production · 12,000 real production prompts mined from 30 days of production logs into an automated evaluation suite with LLM-judge scoring and hallucination tracking · head-to-head evaluation stopped a model swap that looked faster but scored no better · Amazon Bedrock adopted as the unified model channel, with cutover sequenced behind a pre-migration cost baseline
Read the case study →
Identity protected
AWS Well-ArchitectedAWS IoT CoreIoT Device ShadowAmazon RDS
AWS IoT CoreGenerative AI Applications · IoT & Cloud Operations
Production Engagement · Anonymized
Caban hardens a mission-critical IoT energy platform: a 23-finding WAR
Six-pillar-scoped Well-Architected Review · 23 high-risk findings documented · phased console-to-IaC remediation + HTTPS-to-MQTT IoT modernization
Read the case study →