UNDER THE HOOD

Platform architecture

Every Foundry product runs on a shared platform — common orchestration, governance, and infrastructure. One architecture, multiple deployment modes, consistent security posture.

01

InfoTech Foundry Stack

Five layers from application to infrastructure. Products sit at the top; governance runs through every layer.

APPLICATION LAYER Ops Agent ServiceHub DocFlow CX Agent Risk Lens Market Pulse ORCHESTRATION & AGENT RUNTIME Agent Runtime Workflow Engine Tool Registry HITL Gate SOP Engine Cron AI SERVICES LLM Gateway RAG Pipeline ML Models Vector Store NLP / OCR Voice AI GOVERNANCE & COMPLIANCE GuardRail AuditTrail DataVault RegWatch PII Shield RBAC INFRASTRUCTURE KubernetesKafkaRedisPostgreSQLS3 / BlobPrometheusGrafana DEPLOYMENT MODES Cloud AWS · Azure · GCP On-Premise Private DC · Air-Gapped Hybrid On-prem Data Sovereign In-country · Local hyperscaler SAME CODEBASE ALL MODES

Application

Product-specific logic. Each product is a composition of orchestration primitives, AI services, and domain models. Independently deployable.

Orchestration

Agent runtime, workflow engine, tool registry, and human-in-the-loop gates. The brain that coordinates multi-step AI workflows with approval chains.

AI Services

LLM gateway (model-agnostic), RAG pipelines, ML model serving, vector store, NLP/OCR, and voice AI. Shared across all products.

Governance

GuardRail, AuditTrail, DataVault, RegWatch, PII Shield, RBAC. Runs through every layer — not bolted on after the fact.

Infrastructure

Kubernetes-native. Kafka for events, Redis for cache, Postgres for state, S3-compatible storage. Prometheus + Grafana for observability.

02

Request lifecycle

Every request — whether a voice call, API query, or document upload — follows the same governance-wrapped lifecycle.

01

Ingest

API, voice, email, webhook, file upload

02

GuardRail

Auth, PII scan, policy check, rate limit

03

Orchestrate

Route to agent, workflow, or model pipeline

04

Process

LLM call, ML inference, RAG retrieval, tool use

05

HITL Gate

Human review if confidence < threshold or policy requires

06

Validate

Output check, PII re-inject, schema validate

07

Respond

Deliver result, log to AuditTrail, emit event

Governance checkpoints Processing stages

Governance is not a layer — it's a wrapper

Every request enters and exits through GuardRail. Every decision is logged to AuditTrail. This isn't optional — it's structural.

HITL is configurable, not binary

Set confidence thresholds per product, per case type, per user role. High-confidence actions auto-proceed. Low-confidence routes to human review.

Events, not requests

Every step emits events to Kafka. Downstream systems can subscribe — analytics, billing, external notifications, or custom integrations.

03

Deployment models

Same codebase, same API surface, same governance — deployed wherever your data needs to live.

Cloud

Managed deployment on AWS, Azure, or GCP. We handle scaling, patching, and monitoring. You focus on your product.

✓ Auto-scaling

✓ Managed upgrades

✓ Multi-region

✓ 99.95% SLA

On-Premise

Full deployment in your data center. Air-gapped option available. Your hardware, your network, your keys.

✓ Data never leaves your DC

✓ Air-gap capable

✓ BYOK encryption

✓ Custom SLA

Hybrid

Control plane in cloud, data plane on-prem. Get managed orchestration with sovereign data residency.

✓ Best of both

✓ Managed control plane

✓ Data stays on-prem

✓ Flexible scaling

PKSA

Sovereign

In-country deployment on local hyperscalers or government-approved infrastructure. Meets national data residency laws.

✓ Country-level pinning

✓ Local hyperscaler

✓ Regulatory approved

✓ Gov-cloud compatible

04

Integration patterns

Foundry products integrate with your existing systems — not replace them. Every product exposes REST/GraphQL APIs, event streams, and webhook endpoints.

SYNCHRONOUS

REST / GraphQL APIs

Standard request-response for queries, actions, and real-time interactions. Full OpenAPI specs. SDKs for Python, Java, and Node.js.

POST /api/v1/agents/ops/cases

GET /api/v1/intelligence/risk/alerts

POST /api/v1/docs/classify

ASYNCHRONOUS

Event Streams

Kafka topics for every product event — case created, alert triggered, document processed, workflow completed. Subscribe and react.

foundary.ops.case.created

foundary.risk.alert.fired

foundary.docs.extracted

PUSH

Webhooks & Connectors

Pre-built connectors for SAP, Salesforce, ServiceNow, Epic, and 40+ enterprise systems. Custom webhooks for everything else.

SAPSalesforceServiceNowEpicKafkaSMTP

05

Security & compliance

Security is structural, not a feature. Every layer of InfoTech Foundry stack is built with zero-trust principles, encryption at rest and in transit, and comprehensive audit logging.

Encryption

AES-256 at rest. TLS 1.3 in transit. Field-level encryption for sensitive data. BYOK supported with HSM integration.

Identity

SAML 2.0, OIDC, LDAP. Attribute-based access control (ABAC). MFA enforced. Session management with configurable timeouts.

Compliance

ISO 27001. SOC 2 Type II. HIPAA. GDPR. PCI DSS ready. EU AI Act high-risk system logging. SECP and SBP frameworks.

Audit

Every AI decision logged with inputs, outputs, model version, confidence, and reasoning chain. Tamper-proof hash chains. Configurable retention.

READY TO GO DEEPER?

Request a technical deep-dive.

Our engineering team will walk through the architecture with your technical stakeholders — covering deployment, security, integration, and governance requirements.