Agentic Process Orchestration Platform

Your AI-powered foundation for governance, risk, and compliance solutions. Built on AWS EKS with enterprise-grade architecture designed for rapid tenant deployment and scale-to-zero cost optimization. Built with our Agentic SDLC Platform.

3
Accelerator Demos
4
Kubernetes Namespaces
<1 day
Tenant Deployment
$76/mo
Idle State Cost*

AWS Account Foundation

Your cloud infrastructure runs on a structured AWS Organization with Control Tower governance and Identity Center for secure access management.

Account Details

Account ID
••••••••9416
Primary Region
us-east-1
Availability Zones
us-east-1a, 1b, 1c
VPC CIDR
10.0.0.0/16
Environment
Demo / Non-Production

AWS Organizations Hierarchy

Master Billing Account

Consolidated billing with cost allocation tags. All member accounts roll up to a single invoice with detailed usage breakdowns by service and namespace.

Cost Management

AWS Control Tower

Landing zone with guardrails enforcing security baselines. Preventive and detective controls monitor compliance across all accounts.

Governance

IAM Identity Center (SSO)

Federated access using AWS SSO with profile-based authentication. The account-creation profile provides administrator access through AWSReservedSSO_AWSAdministratorAccess role.

Access Management

APOP Demo Account

Isolated workload account hosting the EKS cluster and supporting services. Designed for demonstration and development activities.

Workload

EKS Cluster Architecture

A Kubernetes 1.31 cluster optimized for cost efficiency through SPOT instances and scale-to-zero capabilities. Full Terraform automation enables rapid teardown and rebuild.

Component Configuration Details
Cluster Name apop-eks Kubernetes v1.31 (supported until Nov 2025)
Node Group SPOT Only Mandatory t3.medium, t3a.medium, t3.large, t3a.large
Scaling 0 - 6 nodes Min: 0 (scale-to-zero) | Desired: 3 | Max: 6
Storage gp3 EBS Volumes 50GB per node, KMS encrypted, 3000 IOPS
Network Public Subnets NAT Gateway disabled (saves $32.85/mo) 1
Endpoint Access Public + Private API server accessible from both networks

Cluster Add-ons and IRSA Roles

🔌

CoreDNS

Kubernetes DNS resolution for service discovery

🌐

VPC-CNI

Native AWS networking with pod IP addressing

💾

EBS CSI Driver

Persistent volume provisioning with IRSA

⚖️

Cluster Autoscaler

Automatic node scaling based on demand

🔀

Load Balancer Controller

AWS ALB/NLB integration for ingress

🚀

Karpenter (Optional)

Advanced node provisioning and scheduling

📦

Backup Service

S3 backup with least-privilege IRSA role

📊

CloudWatch Logging

API, audit, authenticator, scheduler logs (7-day retention)

Agentic SDLC Platform

The platform that builds platforms. This entire demo infrastructure was designed, implemented, and deployed using our Agentic SDLC Platform with autonomous AI agents and Compliance-Driven Development methodology.

CDD Methodology

Compliance-Driven Development integrates regulatory requirements directly into the development lifecycle. Every feature is traced from requirement to implementation with automated evidence collection.

  • Requirement-to-code traceability
  • Automated compliance evidence
  • Pre-commit regulatory validation
  • Audit-ready documentation

Self-Building

This infrastructure demonstrates the platform's capability to build itself. Every aspect of APOP - from Terraform modules to Kubernetes manifests to this presentation - was created using the Agentic SDLC.

  • IaC generation with compliance validation
  • Automated testing and deployment
  • Documentation generated from code
  • Continuous compliance monitoring

Specialized AI Agent Categories (41 agents - click for details)

Orchestration & Planning

Development & Quality

Security & Compliance

Testing & Validation

Performance & Analysis

Deployment & Operations

Content & Utilities

Open Standards Observability

OTEL

Built on OpenTelemetry (CNCF Graduated)

Instrumented once. Export anywhere. Zero re-instrumentation when switching vendors.

Our Reference Stack

Phoenix AI

LLM observability

Prometheus

Metrics & alerting

Grafana

Dashboards

Loki

Log aggregation

Your Enterprise Stack

Datadog

OTEL compatible

Splunk

OTEL compatible

New Relic

OTEL compatible

Dynatrace

OTEL compatible

Same instrumentation exports to ANY OTEL-compatible backend. Your existing tools work out of the box.

Multi-Framework Compliance

SOX EU AI Act GDPR HIPAA ISO 27001 NIST 800-53 SOC 2 PCI DSS

Cloudflare Zero Trust Protection

All platform applications are secured behind Cloudflare Access with OTP authentication restricted to @northhighland.com email addresses. The cloudflared tunnel provides secure connectivity without exposing services directly to the internet.

TUNNEL cloudflared Architecture

Tunnel ID: 8cde61f5-e102-4d3d-a4ab-e4b777da8c54
Connection: Outbound-only (no inbound ports)
Protocol: QUIC with automatic failover
  • Kubernetes services never exposed to public internet
  • Encrypted tunnel from cluster to Cloudflare edge
  • Automatic TLS certificate management
  • DDoS protection at Cloudflare edge

ACCESS OTP Authentication

Email Domain Restriction
@northhighland.com only
  • One-time PIN sent via email
  • 24-hour session duration
  • No VPN required for access
  • Audit logs for all access attempts

Protected Applications

🛡️
SNAP
snap.accelerator-demo.com
🛡️
RegRiskIQ
regriskiq.accelerator-demo.com
🛡️
Cost Reports
reports.accelerator-demo.com
🛡️
This Presentation
apop.accelerator-demo.com

Zero Trust Architecture

Every request is authenticated at the Cloudflare edge before reaching your infrastructure. Users verify identity via email OTP, and only @northhighland.com addresses are permitted. The cloudflared tunnel ensures your Kubernetes services remain private while still accessible to authorized team members.

Knowledge Platform

A comprehensive knowledge management architecture powering Policy Assistant across all accelerators. Strategic curation pipelines feed an intelligent knowledge base, exposed through governed APIs with full AI governance instrumentation.

Platform Architecture Stack

Consumer Interfaces
SNAP | RegRiskIQ | STORI | AI Avatar | Voice (36+ languages)
AI Governance Instrumentation
OpenTelemetry Spans → Phoenix Observability | Control Requirements Satisfied
Agentic Query Processing (Multi-Agent Workflows)
Classify → Evaluate → Rewrite → Optimize → Route
RAG Copilot API (Knowledge Factory Service)
GraphRAG | Hybrid Retrieval | Multi-Provider LLM | Knative Serverless
Knowledge Factory (Intelligent Curation Pipelines)
Document Ingestion | Chunking | Embedding | Relationship Extraction | Entity Resolution
Knowledge Base (Dual-Store Architecture)
pgvector
Vector Embeddings
Neo4j
Knowledge Graph

Knowledge Factory

Strategic pipelines that intelligently curate content into the knowledge base, ensuring high-quality retrieval.

1
Document Ingestion
PDF, DOCX, HTML, Web URLs, APIs
2
Intelligent Chunking
Context-aware segmentation with overlap
3
Embedding Generation
Multi-model support (OpenAI, Cohere, local)
4
Relationship Extraction
Entity linking, policy hierarchies, citations
5
Quality Validation
Deduplication, freshness scoring, metadata

Agentic Query Processing

Multi-agent workflows that evaluate, classify, and optimize queries before retrieval and synthesis.

C
Classifier Agent
Intent detection, topic categorization
E
Evaluator Agent
Query quality assessment, ambiguity detection
R
Rewriter Agent
Query expansion, reformulation for retrieval
O
Optimizer Agent
Retrieval strategy selection, ranking
S
Synthesizer Agent
Response generation with citations

AI Governance Instrumentation (click for details)

Every query through the Knowledge Platform is fully instrumented with OpenTelemetry spans, feeding into Phoenix for observability. This directly satisfies AI Governance control requirements for this use case.

OpenTelemetry Spans

  • • Query classification traces
  • • Retrieval latency metrics
  • • LLM call instrumentation
  • • Token usage tracking

Phoenix Observability

  • • LLM evaluation dashboards
  • • Response quality scoring
  • • Hallucination detection
  • • Cost analysis per query

Controls Satisfied

  • • OM-01: Performance monitoring
  • • OM-03: Output validation
  • • GA-02: Prompt management
  • • IM-01: Incident detection

Multi-Provider LLM

OpenAI, Anthropic Claude, Azure OpenAI, AWS Bedrock, Google Vertex AI with automatic failover.

AI Avatar Interface

Video avatars and virtual assistants deliver guidance through conversational AI and training content.

Multilingual Support

36+ languages via Deepgram Nova-2 STT/TTS including Spanish, French, German, Japanese, Korean, Mandarin.

AI Governance Program

Fully codified AI governance playbook with 63 controls across 14 domains. Controls implemented as code with complete regulatory framework mapping to ISO 42001, EU AI Act, and NIST AI RMF.

63
Total Controls
14
Governance Domains
4
Code Formats
4
Regulatory Frameworks

AI Governance Program Foundation (click for details)

5 domains | 19 controls | Enterprise-wide governance structure

GO Governance & Accountability 5 controls
RO Regulatory Oversight 5 controls
TP Third-Party Management 2 controls
CO Communications 2 controls
AA Assessment & Assurance 5 controls

AI Use Case/System Lifecycle (click for details)

9 domains | 44 controls | Per-system governance controls

RM Risk Management 6
LC Lifecycle Mgmt 5
SE Security 6
RS Responsible AI 5
GA Generative AI 6
PR Privacy 4
OM Operations 5
IM Incident Mgmt 3
PL Project Lifecycle Management 4

Regulatory Framework Coverage

ISO 42001:2023
AI Management System - Full Mapping
EU AI Act
Articles 9-26 - Full Mapping
NIST AI RMF v1.0
GOVERN, MAP, MEASURE, MANAGE
ISO 27001 / SOC 2
Security Controls - Integrated

Controls as Code Implementation

📄
Markdown
63 control directories with best practices, evidence requirements, roles
🐍
Python
Structured control_definitions.py with automation levels
⚛️
TypeScript
Interactive AIGovernanceControlDomains UI component
🔗
Neo4j
Graph relationships with control-to-framework crosswalk

Accelerator Demos

Three specialized accelerators address distinct governance domains. Each runs in an isolated Kubernetes namespace with dedicated databases and resource quotas. Currently scaled to zero for cost optimization.

SNAP

Policy Intelligence Platform

Core Components

  • Camunda BPM workflow orchestration
  • Flask backend API (port 5003)
  • React/Next.js frontend UI
  • Neo4j graph visualizer
  • PostgreSQL (snap_db)
  • Redis session/cache layer

Use Cases

  • Benefits policy management
  • Policy document governance
  • Workflow-driven approvals
  • Graph-based policy relationships

RegRiskIQ

Regulatory Compliance Platform

Core Components

  • Camunda BPM with full authorization
  • FastAPI backend (port 8000)
  • Next.js frontend (port 3000)
  • RAG Copilot with GraphRAG
  • Neo4j 5.15 Enterprise + APOC + GDS
  • PostgreSQL with pgvector embeddings

Use Cases

  • SOX/GDPR compliance tracking
  • AI-powered risk analysis
  • Regulatory document search
  • Real-time risk scoring

STORI

Data Governance Platform

Core Components

  • Camunda BPM with init containers
  • OPA policy enforcement (2 replicas)
  • Trino SQL query engine
  • Data governance application
  • Neo4j for data lineage graphs
  • MinIO object storage (S3 backed up)

Use Cases

  • NIST framework compliance
  • Data lineage tracking
  • OPA policy enforcement
  • SQL federation via Trino

Knowledge Factory Supporting Namespace

Enterprise document processing and knowledge extraction service. Ingests documents, processes content, and builds relationship graphs for all platform solutions.

API: Document ingestion (port 8000)
Database: PostgreSQL
Graph: Neo4j relationships
Cache: Redis processing queue

Shared Platform Services

The apop namespace hosts shared infrastructure components that all solutions consume. These stateful services provide the foundation for AI capabilities and data persistence.

Service Version Configuration Purpose
PostgreSQL pgvector/pg15 20Gi storage, 4 databases (apop, camunda, content, auth) Relational data with vector embeddings
Neo4j 5.15 Enterprise APOC + GDS plugins, 1Gi heap, 20Gi storage Knowledge graphs and relationship queries
Redis 7-alpine AOF persistence, 512MB max, 10Gi storage Caching and session management
Kafka + Zookeeper 7.4.0 3 partitions, 168-hour retention Event streaming and async processing
OPA Latest 2 replicas, ConfigMap policies Policy-as-code enforcement
RAG Copilot v2.0 Knative scale 0-10, 50 concurrent, 5min timeout GraphRAG search with multi-hop reasoning

MCP Agent Platform (Knative Serverless)

Dispatcher

Port 8190

Routes incoming requests

Classifier

Port 8191

AI-powered classification

Router

Port 8192

Request routing logic

Sender

Port 8193

SMTP notifications

Process Automation Assets

Executable business processes and decision models across all accelerators. Built on Camunda Platform 7 with full BPMN 2.0 and DMN 1.3 compliance.

180
BPMN Process Models
Click to explore catalog
35
DMN Decision Tables
Risk, compliance & governance decisions

Process Models by Accelerator

91 core processes shared across accelerators

STORI

81 processes AI governance focus

RegRiskIQ

178 processes Regulatory & risk focus

Knowledge Factory

Shared core Content pipeline focus

SNAP

Shared core Learning focus

Control-to-Process Mapping

Each of the 63 AI Governance controls maps to one or more executable BPMN processes, providing automated workflow enforcement and evidence collection.

GA: Governance RM: Risk DM: Data ML: Lifecycle OM: Operations

Platform Capabilities

Complete demo-ready infrastructure in minutes. Strategic fixtures, automated testing, and full backup/restore capabilities enable rapid tenant deployment with zero manual data entry.

Strategic Fixtures System

Spin up solution AND all required data instantly. Per-namespace fixtures with schema validation and CDD integration.

/fixtures/
snap/ | regriskiq/ | stori/ | knowledge-factory/
00-schema.cypher | seed-data.sql | synthetic-data.json
  • ✓ Zero manual data entry for demos
  • ✓ Consistent, repeatable test scenarios
  • ✓ Synthetic data (no PII concerns)

Full Backup & Restore

Per-namespace recovery with automated S3 backup. Full cluster restoration in ~15 minutes.

backup.sh
Timestamped backups
restore.sh
Full namespace restore
load-fixtures.sh
Synthetic/demo data
restore-cluster.sh
Full cluster (~15 min)
S3 Bucket: apop-eks-backups-••••••••9416

E2E Test Automation

Playwright-based testing from Agentic SDLC Platform with fixture-based data injection.

  • playwright-e2e-tester SubAgent
  • test-coverage-analyzer SubAgent
  • ▶ Full UI path coverage
  • ▶ CDD compliance test scenarios

{} Infrastructure as Code (100%)

Fully reproducible infrastructure. Destroy and recreate identically in minutes.

Terraform
EKS, Cloudflare, DNS
K8s Manifests
Declarative services
Helm Charts
Standardized deploys
GitOps Ready
ArgoCD/Flux compatible

Event Streaming: Kafka (Implemented, Not Deployed)

Kafka event streaming capability is fully implemented in the APOP platform but not currently deployed in production solutions. Available for future environments requiring async event processing.

Security Posture (Demo Environment)

This non-production environment implements baseline security controls aligned with SOC2 Type 2 requirements. Production deployments require additional hardening measures.2

Encryption

  • At Rest: KMS keys for EBS, S3, and Kubernetes secrets
  • EBS Volumes: gp3 with customer-managed KMS key rotation
  • S3 Backups: AES256 server-side encryption
  • Secrets: Kubernetes secrets encrypted via cluster KMS

Access Control

  • IMDSv2: Required on all EC2 nodes (SSRF protection)
  • IRSA: Service-level IAM roles for workloads
  • SSO: AWS Identity Center with federated access
  • OPA: Policy-as-code runtime enforcement

Network Security

  • VPC: Isolated 10.0.0.0/16 network
  • Security Groups: Inbound restricted to VPC CIDR
  • Cloudflare Access: Email domain restrictions (northhighland.com)
  • API Endpoint: Public + Private access enabled

Audit and Logging

  • CloudWatch: API, audit, authenticator, scheduler logs
  • Retention: 7 days (cost-optimized for demo)
  • Backups: SHA256 checksums for integrity
  • Compliance Tags: SOC2-Type2 labels on all resources

Multi-Tenancy Architecture

4 tenants, 1 platform, complete isolation. Each accelerator runs in fully isolated environments with dedicated databases and resource quotas.

Isolation Layers

K8s
Kubernetes Namespace

Namespace-per-tenant with resource quotas

PG
PostgreSQL Schema

Schema-per-tenant for relational data

N4j
Neo4j Database

Database-per-tenant for graph data

R
Redis Cache

Namespace-per-tenant with key prefixing

CF
Network Access

Cloudflare Access policies per tenant

Cost Allocation

$19
per tenant / month (idle)
Total Idle Cost $76/month
Number of Tenants 4
Cost per Tenant $19/month
Active demo mode: ~$70/tenant/month

Production Environment Requirements

The following security measures would be implemented before hosting customer data or providing capabilities beyond demonstration purposes.

1 Network Hardening

  • Enable NAT Gateway for private subnet node placement
  • Implement AWS PrivateLink for service endpoints
  • Deploy AWS WAF with OWASP rule sets
  • Configure VPC Flow Logs with extended retention
  • Enable AWS Shield Advanced for DDoS protection
  • Implement network segmentation with dedicated VPCs per tenant

2 Access and Identity

  • Integrate with enterprise IdP (Okta, Azure AD, Ping)
  • Implement MFA enforcement for all console access
  • Deploy AWS Secrets Manager for credential rotation
  • Enable Session Manager for bastion-less access
  • Implement just-in-time access provisioning
  • Configure break-glass procedures with audit trails

3 Data Protection

  • Enable cross-region backup replication
  • Implement point-in-time recovery for databases
  • Deploy AWS Backup with compliance policies
  • Configure data classification tagging
  • Implement customer-managed KMS keys per tenant
  • Enable S3 Object Lock for immutable backups

4 Monitoring and Compliance

  • Enable AWS Config with conformance packs
  • Deploy AWS Security Hub with CIS benchmarks
  • Implement GuardDuty threat detection
  • Configure CloudTrail with log file validation
  • Extend log retention to 365+ days
  • Enable AWS Audit Manager for compliance evidence

5 High Availability

  • Deploy across 3+ Availability Zones
  • Implement multi-region failover capability
  • Configure RDS Multi-AZ for database HA
  • Deploy Redis cluster mode with replication
  • Implement pod disruption budgets
  • Configure horizontal pod autoscaling

6 Operational Excellence

  • Implement GitOps with ArgoCD or Flux
  • Deploy service mesh (Istio/Linkerd) for mTLS
  • Configure PagerDuty/OpsGenie integration
  • Implement chaos engineering testing
  • Deploy Velero for cluster backup/restore
  • Configure automated compliance scanning
Demo Environment Notice:

This environment is designed for rapid iteration and cost efficiency. Infrastructure can be torn down and rebuilt on demand using ./scripts/demo-control.sh. Scale-to-zero capability reduces idle costs to ~$76/month (EKS control plane only). Full destruction reduces monthly cost to $0.

Cost Optimization Strategy

Aggressive cost management through SPOT instances, scale-to-zero Knative services, and automated scheduling.

State Monthly Cost Configuration Use Case
Active Demo ~$279/month 3 SPOT nodes running, all services active Client demonstrations, development
Idle (Scaled Down) ~$76/month 0 nodes, EKS control plane only Weekends, overnight, extended idle periods
Destroyed $0/month Infrastructure fully removed via Terraform Long-term idle, budget constraints

demo-control.sh start

Scales nodegroup to 3 nodes. Environment ready in 3-5 minutes.

demo-control.sh stop

Scales nodes to zero. EKS control plane continues running.

demo-control.sh destroy

Full Terraform destroy. Requires confirmation typing "DESTROY".

AI/MLOps Maturity

AI operations designed for enterprise scale with multi-provider support, cost optimization, and responsible AI practices.

AI Avatar Support From TNBizBot

Integrated AI-powered virtual assistants and video avatars for interactive training, conversational agents, and enhanced user engagement across all accelerator demos.

✓ Virtual Assistants ✓ Video Avatars ✓ Training Content ✓ Conversational AI

Multi-Provider LLM

  • ● OpenAI GPT-4
  • ● Anthropic Claude
  • ● Azure OpenAI
  • ● AWS Bedrock
  • ● Google Vertex AI

Cost Optimization

  • ● Redis query caching
  • ● Semantic deduplication
  • ● Token usage tracking
  • ● Per-request cost metrics
  • ● Phoenix AI observability

Responsible AI

  • ● PII redaction
  • ● Prompt injection protection
  • ● Response verification
  • ● Citation to sources
  • ● gemini-judge validation

Disaster Recovery

15-minute recovery, enterprise-grade resilience with automated backup and restore procedures.

15 min
RTO (Recovery Time)
Verified via restore-cluster.sh
1 hour
RPO (Recovery Point)
Hourly backup schedule
30 days
S3 Retention
Backup retention period
3 AZs
Availability Zones
us-east-1a, 1b, 1c

Extensibility & Integration

Integrates with your existing tools and enables custom development through open APIs.

APIs

  • ● GraphQL for all services
  • ● REST endpoints
  • ● OpenAPI documentation
  • ● Webhook notifications

Integrations

  • ● Slack notifications
  • ● Email alerts (SMTP)
  • ● Jira integration
  • ● Azure DevOps

Custom Development

  • ● MCP protocol for agents
  • ● Agentic SDLC SDK (npm)
  • ● SubAgent framework
  • ● Custom fixtures