Full-Stack Architect · AI Systems · Tel Aviv

SystemsthatScale+

nir.lichtenstein — about.tsavailable

My name is Nir LichtensteinNir Lichtenstein I build AI-driven platforms from first principles — multi-tenant SaaS, RAG pipelines, international SEO infrastructure, complex permission architectures. Twenty years making systems that outlast their requirements.

<Technical Scope/>
Stack
ReactNext.jsTypeScriptPrismatRPCNode.js
AI
RAG PipelinesVector EmbeddingsPrompt EngineeringOpenAI
Architecture
Multi-tenant SaaSRBACWebhooksGeo-targeting
SEO
InternationalHreflangStructured DataCountry-scoped URLs
01Expertise

What I Actually Build

Not features. Systems with deliberate architecture, documented trade-offs, and long-term maintainability — designed from day one.

module_01

AI Systems & Pipelines

Multi-stage prompt architectures, RAG engines with internal memory schemas, structured content generation with evaluation loops and schema-constrained outputs.

OpenAIRAGEmbeddingsPrompt Eng.
module_02

High-Scale SaaS Architecture

Multi-tenant platforms with complex permission modeling, modular content engines, webhook systems, and scalable data flows.

Next.jstRPCPrismaRBAC
module_03

International SEO Infrastructure

Geo-targeting systems, hreflang architectures, country-specific URL strategies. SEO as a first-class engineering concern.

Geo-targetingHreflangi18n
02Flagship System

The Platform That Operates Website Networks

A multi-tenant SaaS platform for managing unlimited websites under a single control plane — organized as Projects → Sites → Markets — with effortless domain swapping, template-driven launches, and API-first delivery.

Beyond a CMS: it includes vector memory, cosine similarity retrieval, and a governed RAG pipeline that builds contextual content and cross-page relationships — without running AI at request time.

“Not a single website — a scalable operating system for brands, domains, and markets.”

Scaling is a way of thinking at the infrastructure level. This system is built for networks. It separates infrastructure, content, and presentation so each tenant can launch new brands through configuration, keep content consistent through governance, and generate new pages using retrieval-powered memory rather than copy-paste.

Each site can run a unique design system or a shared template pack — while consuming the same API-driven core, the same data layer, and the same versioned content artifacts.

PlayGuidePro is a live, production-grade reference implementation built on this platform. The complete infrastructure — including the control plane, CMS layer, vector memory system, and RAG pipeline — can be reviewed in detail during an interview.

TypeMulti-tenant SaaS for website networks
HierarchyProjects → Sites → Markets
CMSStructured sections + reusable entities
MemoryChunked embeddings (semantic index)
RAGCosine retrieval + governed composition
DeliveryAPI-first, headless, template or custom UI
RuntimeVersioned artifacts · no AI on request
schema_01Platform Topology
[ PLATFORM ]Multi-tenant control plane
PROJECT_APortfolio: website network project
SITE: Brand TenantTheme / template + domain
DOMAIN_APrimary domain mapping
DOMAIN_BSwap / relaunch in minutes
+ NEW_DOMAINConfig-only launch
+ NEW_SITENew brand · template or custom UI
+ PROJECT_BAnother website network
schema_02Network Content Pipeline
TENANT://
Multi-Tenant Control Plane
Projects · Sites · Markets
Tenant Isolation
Config-over-Code
Domain Swaps
The operating layer that manages Projects → Sites → Markets. New brands and domains launch through configuration, without duplicating infrastructure or code.
CMS://
Structured Content System
Sections · Entities · Rules
Composable Blocks
Typed Content Models
Per-Tenant Config
A structured CMS for building pages from reusable blocks and entities. Content is governed per tenant so scale doesn’t degrade quality or consistency.
VEC://
Vectorized Memory Index
Chunks → Embeddings
Chunk Store
Semantic Index
Query-Ready
Selected content chunks are stored as embeddings. This creates a semantic index that powers retrieval, contextual expansion, and intelligent cross-page relationships.
RAG://
Retrieval & Composition
Cosine search → context
Cosine Similarity
Policy Guardrails
Schema Constraints
During page generation the system performs cosine similarity retrieval to find the most relevant chunks and entities, then composes a governed context under tenant policies and schema constraints.
API://
Headless Delivery Layer
Templates · Themes · Custom UI
API-First
Template Systems
Unique Site Design
Sites render via an API-driven, headless layer. Each site can ship a unique design system or reuse templates while consuming the same core infrastructure and content artifacts.
// Traditional approach
One site per codebase (hard to scale brands)
Manual duplication per market / domain
No semantic memory between pages
AI used ad-hoc, not as infrastructure
Updates are risky across thousands of pages
vs
// This platform
Unlimited sites under one control plane
Domains & markets launched by configuration
Vector memory enables semantic reuse & linking
RAG pipeline governed by tenant rules
Versioned artifacts → safe, deterministic delivery
schema_03Request Resolution Lifecycle
ENTRYIncoming RequestEdge Middleware
Edge-first routing — fast, predictable, and designed to avoid unnecessary backend work.
RESOLVETenant / Market ResolutionURL · Headers · Defaults
Resolves which tenant and market should serve the request using deterministic rules (locale routes, headers, and fallbacks).
POLICYTenant GovernanceRules + compliance
ROUTERouting GuaranteesLocale & domain mapping
DBTenant DataEntities + configs
VECVector MemoryChunk retrieval
CACHEEdge CacheTenant+market keys
OUTPUTAPI Response / SSR RenderArtifacts + template/theme
Serves versioned artifacts through API-first delivery. Frontend renders via template systems or unique design, without runtime AI dependency.
03Selected Work

Live Systems

Real platforms, real scale. Each a case study in architecture decisions and engineering depth.

VOL_01VOL_02VOL_03VOL_04MONTUEWEDTHUwebhook01 / 06
Operational SaaS · Crisis Coordination2024

EVP Volunteers

Operational control system for volunteer coordination during large-scale events and emergency scenarios. Includes structured team models (Squads), granular RBAC, event-driven workflows, RESTful API integrations, and real-time dashboards for command-level visibility.

Next.jsREST APIRBACEvent WorkflowsTypeScript
View details
/ca/en/uk/en/au/en/nz/enPLATFORM02 / 05
Affiliate · SEO · AI Infrastructure2024–25

PlayGuidePro

International affiliate platform built on a custom multi-tenant website engine. Market-aware routing, hreflang architecture, governed vector memory, and a contract-first RAG pipeline that generates versioned content artifacts with zero AI at request time.

Next.jsPrismaRAGEmbeddingsRedisGeo/Locale RoutingHreflang
View details
OWNERADMINMANAGEReditorviewereditorviewerRBAC · per-org context · tRPC middleware03 / 06
Internal SaaS · Financial Operations2025

Donor Management

Financial-grade donor operations platform with multi-tier RBAC, transaction integrity, automated email workflows, and webhook-driven integrations for payment processing and reporting.

Next.jsPrismaRBACResendREST API
View details
שאלהLIVE · 3 venues · RTL04 / 06
SaaS · Event Management2023–24

MyQuiz

Event management and live trivia platform supporting ticket configuration, attendee management, event-level permissions, automated registration flows, and a modular content engine.

ReactTypeScriptEvent WorkflowsRBACRTL
View details
SYSTEMARCHITECTURETRADE-OFFSNARRATIVECLARITYSYSTEM → NARRATIVE → ALIGNMENT05 / 06
Meta · Systems Framing2026

Systems Portfolio

A structured technical narrative platform designed to translate complex architectures into executive-level clarity, decision-ready models, and presentation-grade system visibility.

System FramingInformation ArchitectureTechnical NarrativeExecutive Communication
View details
CLIENTBALANCESTATUSACC_4421$24,800ACTIVEACC_7703$11,250PENDINGFOREX · TRADING · CRM06 / 06
FinTech · Trading Infrastructure2021–2023

Forex Trading & CRM Platform

Led end-to-end architecture and delivery of a production Forex trading system and client management platform, overseeing system design, transactional modeling, and cross-functional execution.

System ArchitectureFinTechTransactional WorkflowsCRMTeam Leadership
View details
04Engineering Decisions

Problem → Decision

Not what I built — why I built it that way. The trade-offs that matter.

01
PROBLEMPlayGuidePro

Scaling multi-country content without multiplying infrastructure complexity

✗ Not this

Clone the application per market. Country growth becomes a code management problem.

✓ Decision

Architected a single codebase with a market configuration layer. Countries exist as governed config objects — routing, operators, bonuses, and content derive from structured tenant definitions instead of forks.

ResultMarket expansion reduced from architectural effort to configuration overhead. 3 markets launched in 48 hours without code branching.
02
PROBLEMDonor Management

Evolving permission models without creating regression risk

✗ Not this

Scatter role checks across API handlers. Adding a role means refactoring dozens of endpoints.

✓ Decision

Centralized context-aware RBAC at the router middleware layer. Permissions evaluated per-organization before execution. Role expansion becomes a configuration update — not a refactor.

ResultZero permission regressions after expanding from 3 to 5 role tiers. Authorization logic remains isolated and testable.
03
PROBLEMEVP Volunteers

Operational visibility close to real-time without persistent socket infrastructure

✗ Not this

Introduce a dedicated WebSocket server for all dashboards. High infra overhead for low-frequency updates.

✓ Decision

Implemented webhook-driven event propagation with optimistic UI updates and targeted cache revalidation. State converges within seconds without maintaining open connections.

ResultNear-real-time operational UX at a fraction of infrastructure complexity and maintenance cost.
04
PROBLEMPlayGuidePro · AI Layer

Using AI for content generation without introducing runtime unpredictability

✗ Not this

Generate content dynamically on request. High latency, inconsistent output, and no audit trail.

✓ Decision

Separated generation from delivery. Content is generated through a RAG pipeline, stored as versioned artifacts, and served statically. AI acts as an infrastructure layer — not a runtime dependency.

ResultDeterministic page delivery with AI-assisted creation. No AI execution in production request cycle, full traceability of generated artifacts.
05Philosophy

How I Think

I approach systems as evolving architectures — not isolated features. The goal is always scalability, semantic coherence, and long-term maintainability.

Every decision has trade-offs. Every trade-off should be deliberate. That's what separates engineering from coding.

01
Architecture before code

Architecture mistakes rarely break on day one. They compound silently until change becomes expensive.

02
AI as infrastructure, not magic

A well-defined layer with clear inputs, outputs, and failure modes — not a black box you hope will work.

03
Scalability is a feature

Schema design, permission modeling, and content architecture that grows without rewrites.

04
Full ownership, end to end

From DB schema to pixel-perfect UI. Fewer handoffs, faster iteration, better coherence.

Let's
build
something.

Available for senior engineering roles, architecture consulting, and AI system design. Based in Israel, working globally.