SARAI
System for Analysis & Response AI — Turn Data Into Decisions
A unified AI marketing data platform consolidating data from 600+ sources (Google Ads, Meta Ads, GA4, CRM, e-commerce, etc.) into actionable business insights and automated decisions powered by AI.
Full-Stack Developer (Internship)
I joined this project as a Full-Stack Developer Intern, implementing features according to the high-level architecture and scope defined by leadership. During development, I ended up taking hands-on ownership of the vast majority of end-to-end technical execution:
The Problem
Marketing teams typically operate with fragmented data—each platform (Google Ads, Meta Ads, GA4, CRM) maintains its own dashboard, isolated format, and disconnected ecosystem. Consolidating this data into unified weekly reports required hours of manual spreadsheet manipulation, making real-time insight extraction or anomaly detection nearly impossible.
The Approach & Engineering Execution
How technical challenges were solved and executed hands-on during development.
Airbyte Connector Configuration
Using Airbyte Cloud as the data integration layer was set in the project scope. However, configuring each connector, handling OAuth flows, and debugging unexpected API responses fell directly on my hands.
Covering Failed Team API Scope
Several API integrations initially allocated to other team resources were not completed. To keep the project on schedule, I took over, rebuilt, and delivered those API integrations independently.
Layered Architecture Integration
I implemented each layer according to design specs: Frontend (Next.js) → Backend API (NestJS) → Query & Blend Engine → Data Layer (PostgreSQL + Airbyte) → AI Layer (Claude API).
System Architecture Diagram
Real-time data flow from 600+ ingestion sources to interactive visual dashboards and export engines.
Key Features
8 core modules forming the comprehensive functionality of the SARAI platform.
1. Connect
600+ data connectors (ads, analytics, CRM, e-commerce, databases).
2. Query
Visual query builder supporting custom metrics, dimensions, filters, & scheduling.
3. Blend
Cross-source data join engine supporting custom mathematical formulas.
4. Visualize
Interactive dashboard featuring 8 widget types, drag-and-drop layout, & public links.
5. Analyze
Real-time AI chat companion for trend analysis, anomaly detection, & forecasting.
6. Export
Automated scheduled push to Google Sheets, Excel, Looker Studio, & Power BI.
7. Manage
Multi-tenant team workspace with 4 security role access levels (RBAC).
8. Scale
Subscription & billing infrastructure with active usage tracking.
Tech Stack
Modern tech choices engineered for high throughput, data integrity, and modular scalability.
Process Timeline
Breakdown of the 8 execution phases from zero to a complete, production-ready deliverable.
Outcome & Project Status
SARAI was successfully completed. As a Full-Stack Developer Intern, I initially joined to work on specific features defined in the project overview. However, when certain integration parts planned for other resources stalled, I took ownership of the vast majority of end-to-end technical execution—from building critical APIs to ensuring every system layer (frontend, backend, data, and AI) connected seamlessly and functioned according to design specifications.