Case Studies

Real products powered by
structured data infrastructure.

We turn fragmented Pakistani data into clean APIs, databases, and AI-ready infrastructure. From consumer fintech apps to enterprise AI platforms — here's what that looks like in production.

CardWise

A bank card savings app for Pakistani consumers

Type

API-Powered Product

Industry

Consumer Fintech

Stack

React Native · Supabase

Deals APICards APIMerchants API

cardwise.pk

Challenge

Pakistan has 20+ major banks publishing thousands of card discounts across PDFs, websites, and social media. No standardized format, no unified source. CardWise needed structured, real-time access to every bank deal — filterable by city, category, bank, and card type — without maintaining fragile ingestion pipelines in-house.

What we built

Three production APIs: Deals API(10,000+ active discounts with normalized categories & city filtering), Cards API(full Pakistani card catalog with images & bank info), and Merchants API (enriched merchant data with active deal counts per bank).

How DataLayer handles it

DataLayer manages the full ingestion lifecycle behind the API — automated ingestion, normalization, deduplication, category mapping, validation, and monitoring. Data is refreshed daily from primary bank sources. CardWise simply queries clean, structured endpoints without touching any of the complexity.

10,000+

Active deals structured

20+

Banks covered

Cities

Zero

Ingestion pipelines maintained by CardWise

Internal Build Note

CardWise went from product idea to production app without building its own ingestion infrastructure. All data collection, normalization, and monitoring is handled by the DataLayer platform — the product team focuses entirely on the consumer experience.

Kobin AI

AI-Powered Agency Operating System

Type

Full Custom Build

Industry

Enterprise SaaS / AI

Stack

Next.js 16 · Supabase · Groq

Custom Data Infrastructure

kobin.team

Challenge

Agency founders were drowning in 6+ tools — Slack for chat, Notion for docs, HubSpot for CRM, Linear for tasks. No shared context between them. The AI couldn't see across tools. Every context switch slowed execution and buried important client context. They needed a single, AI-native workspace built from scratch.

What we built

A production-grade SaaS replacing 6 tools in one workspace. Next.js 16, React 19, Supabase with RLS for multi-tenancy. We architected the AI engine, data models, real-time systems, Google Workspace integrations, and Safepay payment infrastructure from a blank repo to deployed product.

6 → 1

Tools replaced

37+

Components shipped

AI tools in agent loop

API route groups

Technical Scope — What We Engineered

AI Engine

MCP-Style AI Agent

14-tool agentic loop. Creates tasks, updates CRM, searches vault, sends messages — all from ⌘K.

RAG / Search

Vault RAG Pipeline

Two-stage semantic search with pgvector: doc-level anchoring → chunk retrieval with MMR diversity.

AI Automation

Meeting Transcription

Desktop recorder → Groq Whisper → auto-tasks, vault decision notes, CRM pipeline updates.

Collaboration

Real-Time Inbox

Project channels, group DMs, file attachments, emoji reactions, Gmail thread integration.

CRM

CRM & Pipeline

7 deal stages, Kanban board, CSV import, auto-lead detection from Gmail, AI stage suggestions.

Intelligence

Workspace Intelligence

Risk detection, bottleneck analysis, team workload, priority scoring — cached via Redis.

Files

Google Drive Vault

Role-scoped folders synced to Drive. Server-side OCR with pdf-parse and mammoth. AI auto-labeling.

Multi-Tenancy

Client Portal

Scoped workspace per client: inbox, tasks, calendar, vault. One-click portal creation.

Payments

Payment System

Safepay PKR integration. Three-tier plan guards enforcing limits on projects, members, and AI.

Full Technology Stack

Layer	Technologies
Frontend	Next.js 16, React 19, TailwindCSS 4, Radix UI, TipTap, D3.js, Recharts
Backend	Next.js App Router, Supabase (PostgreSQL + RLS + Realtime), Upstash Redis
AI / ML	Groq SDK, HuggingFace BGE-small, pgvector, Whisper
Integrations	Google OAuth, Gmail API, Drive API, Calendar, Meet, Web Push
Payments	Safepay (PKR), Plan-gated feature guards
Infra	Vercel, Supabase Cloud, Electron (desktop recorder)

Internal Build Note

Kobin AI was architected and built end-to-end by DataLayer — the AI engine, RAG pipeline, real-time collaboration layer, Google Workspace integrations, and payment infrastructure. Delivered from a blank repo to a deployed, multi-tenant production product.

WebEngage Automation

CRM campaign reporting, fully automated

Type

Process Automation

Industry

MarTech / CRM

Stack

Python · BigQuery · Supabase

Reverse-Engineered APIsIMAPBigQuery

Full engineering breakdown

Challenge

A marketing team ran 500+ campaigns a month on WebEngage — a CRM with no public API for the campaign content and bulk reports they depended on. Every cycle, an analyst spent 40+ hours logging in, exporting, waiting on emails, opening campaigns one-by-one, and hand-mapping journeys to categories.

What we built

A scheduled pipeline that reverse-engineers WebEngage's own dashboard APIs, intercepts its report emails over IMAP, unwraps the download link, categorizes every journey, and lands clean data in BigQuery and a Supabase content repository — then grew into an ops console with an AI assistant.

~40 hrs

Manual work removed / month

500+

Campaigns reported / month

Channels unified

Zero

Manual steps after trigger

Deep Dive

Read how the pipeline bridges an API, an inbox, and cloud storage — API archaeology, four-mode session auth, the email side-channel, and idempotent warehouse loads.

Read

Foodpanda Pipeline

A national restaurant & menu warehouse

Type

Data Engineering

Industry

Food Delivery / Market Data

Stack

Python · PostgreSQL

API Reverse-EngineeringSitemapPostgreSQL

Full engineering breakdown

Challenge

Pakistan's richest restaurant dataset — menus, prices, cuisines, locations — is locked inside one app, served by an undocumented API, gated on delivery location, and sat behind bot protection. There was no way to query it in bulk.

What we built

A pipeline that discovers 29,000+ restaurants from Foodpanda's sitemap, reverse-engineers the internal vendor API — discovering that omitting coordinates bypasses the geo-gate — and lands full menus in an 11-table PostgreSQL warehouse.

29,115

Restaurants discovered

150K+

Menu items ingested

Warehouse tables

Zero

Bot blocks in logged runs

Deep Dive

Read how the pipeline discovers a whole country, slips past the delivery-range gate, normalizes menus into 11 tables, and survives a multi-hour crawl.

Read

FBR Tax Verification

A captcha-gated portal, turned into one call

Type

API Reverse-Engineering

Industry

GovTech / KYC

Stack

Node.js · Zero deps

IRIS Reverse-EngineeringCNIC / NTNBulk CLI

Full engineering breakdown

Challenge

FBR publishes every taxpayer's active-filer status — but only through the IRIS portal, behind what looks like a captcha and a browser-only screen. KYC and compliance teams need to verify a list of CNICs or NTNs, not click through one at a time.

What we built

By reading IRIS's own Angular bundle, the captcha turned out to be client-side only — never checked by the server. The result is a zero-dependency verification client plus a concurrent bulk runner that turns a spreadsheet of IDs into clean, structured results.

1 call

Verification per request

External dependencies

ID types supported

Concurrent bulk workers

Deep Dive

Read the reverse-engineering trail — network capture, the token hiding in the frontend, the captcha that never mattered, and rebuilding the request from scratch.

Read

Deals Sync Pipeline

The engine keeping bank offers fresh daily

Type

Data Engineering

Industry

Fintech / Market Data

Stack

Python · Supabase

API Reverse-EngineeringIncremental SyncPostgreSQL

Full engineering breakdown

Challenge

A deals product is only as good as its last sync. Across 7 banks and 33 cities, tens of thousands of offers each have their own lifecycle — new merchants appear, discounts shift, deals expire. Re-downloading everything nightly is slow and wasteful.

What we built

An incremental sync that reverse-engineers the deals SDK each bank embeds, fetches only what changed via a signature cache, upserts into a Supabaseschema keyed on each offer's URL, and refreshes CardIntel rankings when it finishes.

Banks in one pipeline

9,500+

Live offers maintained

1,200+

Merchants normalized

Cities covered

Deep Dive

Read how the sync speaks the portal's own SDK, skips everything that didn't change, survives interruption, and expires dead offers instead of deleting them.

Read

Data infrastructure you can trust

DataLayer is not an agency. It's a Pakistan-specific data infrastructure company that also builds custom AI and data systems.

Data Freshness

Bank deals refreshed daily from primary sources
Real-time FBR verification using official public lookup sources
Merchant data updated weekly with change detection

Data Quality

Normalized merchant / category / city schema
Automated deduplication and validation pipeline
Structured JSON — no PDFs, no manual parsing

Reliability & Support

Monitored infrastructure with uptime commitments for enterprise clients
Dedicated integration support and onboarding
Custom SLAs and on-premise deployment available

Sample API Response

GET /v1/deals?city=karachi&bank=hbl

{
  "status": "success",
  "data": [{
    "merchant": "Kababjees",
    "bank": "HBL",
    "card": "HBL Debit Card",
    "discount": "25% off",
    "category": "Dining",
    "city": "Karachi",
    "valid_until": "2026-07-31",
    "branches": ["Clifton", "Gulshan"]
  }],
  "pagination": {
    "page": 1,
    "per_page": 20,
    "total": 1284
  },
  "meta": {
    "request_id": "req_abc123",
    "response_time_ms": 42
  }
}

Public-source data, responsibly collected

All data is sourced from publicly available bank websites, government portals, and official merchant listings. We do not access private or restricted systems. FBR verification uses official public lookup channels.

Two ways to work with us

Whether you need structured data delivered to your product or a full custom system built from scratch.

DataLayer Data Solutions

Integrate our structured datasets — Deals, Cards, Merchants, FBR Tax Verification. We handle ingestion, normalization, and monitoring. You focus on your product.

Like CardWise — they consume our data and focus on their app.

Custom Data Infrastructure

Full systems built from scratch — databases, AI pipelines, real-time systems, integrations, and deployment.

Like Kobin AI — we built the entire product end-to-end.

Let's build something real

Whether it's structured data for your product or a full custom platform — we ship production code, not decks.

Get in Touch

Real products powered bystructured data infrastructure.

CardWise

Challenge

What we built

How DataLayer handles it

Kobin AI

Challenge

What we built

MCP-Style AI Agent

Vault RAG Pipeline

Meeting Transcription

Real-Time Inbox

CRM & Pipeline

Workspace Intelligence

Google Drive Vault

Client Portal

Payment System

WebEngage Automation

Challenge

What we built

Foodpanda Pipeline

Challenge

What we built

FBR Tax Verification

Challenge

What we built

Deals Sync Pipeline

Challenge

What we built

Data infrastructure you can trust

Data Freshness

Data Quality

Reliability & Support

Sample API Response

Public-source data, responsibly collected

Two ways to work with us

DataLayer Data Solutions

Custom Data Infrastructure

Let's build something real

Real products powered by
structured data infrastructure.