/ Frontend

Modern, accessible, internationalized. Multi-language UIs (EN, ES, PT, FR, DE), full theming, real-time state, and embedded data viz — one coherent design system.

Framework

React 18 + TypeScript

Build & tooling

Vite 5

State & data

TanStack Query v5, Zustand

UI components

shadcn/ui + Radix UI

Styling

Tailwind CSS v4

Forms & validation

React Hook Form + Zod

Internationalization

i18next — 5 languages in production

Data visualization

Recharts, D3

Animations

Framer Motion

/ Backend

TypeScript end-to-end, Zod validation at every API boundary, modular domain-organized service structure.

Runtime

Node.js 20 + TypeScript 5

Framework

Express.js — modular, domain-organized

Validation

Zod, end-to-end across all endpoints

Sessions & auth

Passport, OIDC/SSO, JWT, custom dual-gate RBAC

Security

Helmet, rate limiting, input sanitization, audit logging

API documentation

OpenAPI / Swagger

/ Data Layer

Polyglot persistence — each engine chosen for the access pattern it serves: relational, vector, graph, document, cache.

PostgreSQL

Primary relational — transactional data, tenant isolation

pgvector + HNSW

Vector embeddings and ANN search for RAG

Neo4j

Graph database for multi-hop knowledge reasoning (Graph RAG)

Redis

Distributed caching, rate limiting, cross-instance coordination

Firestore

Document storage and real-time data sync

BigQuery

Analytics and large-scale data processing

/ AI & Knowledge Layer

All LLM providers routed through an internal AI Gateway: intelligent routing, fallback, cost controls, per-workspace token budgets. No single-provider dependency.

LLM providers

Anthropic Claude, OpenAI GPT-4, Google Gemini, Mistral

Embeddings

OpenAI, Cohere, Voyage

RAG

Hybrid vector + BM25 with custom chunking and metadata enrichment

Graph RAG

Multi-hop reasoning over connected knowledge structures

Agent protocol

MCP — production server exposing KB resources and tools

AI Gateway

Internal control plane — routing, fallback, cost, prompt registry

/ Integrations

Knowledge sources

Zendesk, Confluence, SharePoint, Notion, Google Drive, web scraping, file/image/video ingestion

Outbound sync

Zendesk, Freshdesk, Intercom, Zoho Desk, Gorgias, HubSpot, GLPI, Ivanti, Aranda, BMC Helix

Communication

Slack, Microsoft Teams, email

Object storage

Google Cloud Storage (primary), AWS S3, Azure Blob

Billing

Stripe

/ Observability & Quality

Telemetry

OpenTelemetry — metrics, traces, structured logs with correlation IDs

Testing

Vitest, Testing Library, Supertest, in-memory PostgreSQL for DB-boundary tests

CI/CD

GitHub Actions — typecheck, tests, security scanning, CodeQL, tenant isolation, migration safety

Job processing

Durable PostgreSQL-backed job queue with retry, DLQ, concurrency control

/ Compute

Right compute primitive per workload — serverless for most APIs, containers for complex workloads, managed clusters when scale demands it.

Cloud Run

Primary production target — containerized APIs, web apps, workers. Scales to zero.

Cloud Functions (2nd gen)

Event-driven functions for webhooks, triggers, lightweight integrations

Google Kubernetes Engine

Workloads requiring fine-grained orchestration and autoscaling at scale

Cloud Run Jobs

Batch processing — ingestion pipelines, scheduled exports, data migrations

/ Data & Storage

Cloud SQL (PostgreSQL)

Primary managed relational — automated backups, failover, read replicas

AlloyDB for PostgreSQL

Sub-10ms read latency alongside analytics

Cloud Spanner

Multi-region with strict transactional consistency

Firestore

Document storage and real-time sync with offline-capable SDKs

BigQuery

Enterprise analytics, product metrics, large-scale processing

Bigtable

High-throughput time-series, event logs, telemetry at scale

Memorystore (Redis)

Managed Redis — caching, sessions, distributed rate limiting

Cloud Storage

Object storage — uploads, artifacts, exports, data lake staging

/ AI & Machine Learning

Integration depth between Vertex AI, Gemini, and the rest of GCP eliminates latency, auth, and compliance friction that multi-cloud AI architectures introduce.

Vertex AI

Managed ML — training, fine-tuning, evaluation, registry

Gemini API (via Vertex AI)

Text, multimodal, function calling, long-context reasoning

Vertex AI Search

Enterprise RAG-capable semantic search with built-in grounding

Vertex AI Agent Builder

Conversational AI infrastructure with tool use and multi-turn memory

Document AI

Intelligent extraction — PDFs, forms, invoices, contracts

Natural Language AI

Entity extraction, classification, sentiment in production pipelines

Speech-to-Text / TTS

Voice I/O for conversational and contact center integrations

Translation AI

Multilingual support — LATAM and global enterprise deployments

Vision AI

Image processing, OCR, content moderation

/ Security & Compliance

Full GCP security stack from day one — not added post-launch.

IAM

Least-privilege — every service account has exactly the permissions it needs

Workload Identity Federation

Keyless authentication from CI/CD — no long-lived secrets

Secret Manager

Centralized secrets with versioning, audit logging, rotation

Cloud KMS

Customer-managed encryption keys across SQL, Storage, BigQuery, Pub/Sub

Security Command Center

Unified vulnerability, misconfiguration, threat posture

Cloud Armor

DDoS protection and WAF on every public-facing deployment

Binary Authorization

Only attested, verified container images reach production

VPC Service Controls

Data exfiltration prevention — perimeters around sensitive services

Cloud Audit Logs

Immutable audit trail across all services

/ DevOps & Delivery

Cloud Build

Managed CI/CD — build, test, deploy pipelines

Artifact Registry

Private container/package registry with vulnerability scanning

Cloud Deploy

Progressive delivery — canary, staged rollouts, automated rollback

/ Observability & Operations

Cloud Monitoring

Metrics, SLOs, uptime checks, dashboards

Cloud Logging

Structured JSON logs with correlation IDs

Cloud Trace

Distributed tracing across services and AI pipeline stages

Cloud Profiler

Continuous production profiling

Error Reporting

Automatic exception grouping and alerting

Pub/Sub

Managed event streaming and message bus

Cloud Tasks + Scheduler

Async task queues and managed cron for background workloads

/ Data Pipelines & Analytics

Dataflow

Batch and streaming transformations — ETL and embedding generation at scale

Cloud Composer (Airflow)

Workflow orchestration for complex data pipelines

Looker Studio

Product analytics dashboards and operational reporting

BigQuery

Central analytics layer — cross-product, cross-tenant