CAT Review Platform

• Category: EdTech / AI / SaaS Product

• Timeline: December 2025

• Product Type: IOL Own Product (via 1Punch Inc.)

The Problem

The CAT (Certified Accounting Technician) program, administered by the National Institute of Accounting Technicians (NIAT), represents over 30,000 accounting professionals in the Philippines. Despite this sizable community, the exam preparation landscape remains firmly rooted in traditional methods — classroom-based review sessions, printed handouts, and generic quiz applications.

For working professionals and students outside Metro Manila, accessing quality review materials means traveling to urban centers or relying on outdated self-study resources. The market lacks a platform that combines AI-driven personalized learning with content specific to Philippine accounting standards. The gap is clear: accounting technician examinees need affordable, accessible, and intelligent exam preparation that meets them where they are.

Our Solution

Through 1Punch Inc. — IOL's sister company and a NIAT-accredited CAT training provider since 2017 — we built an AI-native exam preparation platform with Anthropic's Claude API at its core. Rather than bolting AI features onto a traditional quiz app, we designed the platform around Retrieval-Augmented Generation (RAG) from the ground up.

The platform ingests review materials through an automated PDF processing pipeline, converts them into vector embeddings stored in PostgreSQL with pgvector, and uses semantic search to power an AI tutor that provides context-aware coaching. When a student asks a question, the AI doesn't generate a generic response — it retrieves the most relevant passages from actual study materials and synthesizes accurate, exam-specific guidance.

A white-label architecture was designed from day one, allowing the same core engine to serve multiple professional exams through separate branded instances — a decision that has already proven its value with the CPA Review Platform expansion.

Tech Stack

Key Features

• AI Tutor with RAG: Context-aware exam coaching that understands Philippine accounting standards and references actual study materials rather than generating generic advice.

• Semantic Search: Students can search across all uploaded materials using natural language queries, powered by pgvector similarity matching.

• PDF Processing Pipeline: Automated content ingestion converts review PDFs into searchable, AI-accessible vector embeddings — eliminating manual data entry.

• Level-Based Subscriptions: Tiered access matching the CAT program's three examination levels (Certified Bookkeeper, Registered Cost Accountant, Certified Accounting Technician).

• White-Label Architecture: Designed for multi-exam scalability, enabling expansion to other PRC-administered licensure exams.

Results & Impact

Completed. The platform is live with a few hundred registered users. The white-label architecture was confirmed when the same engine was deployed for CPA exam preparation at cpareview.ph.

Key Takeaways

1. RAG architecture turns generic AI chatbots into domain-specific tutors by grounding responses in actual study materials.

2. Building white-label from day one saved months of re-engineering when expanding to the CPA exam.

3. A small team with domain expertise can build a sophisticated EdTech platform without enterprise-level resources.

FAQ

Q: What is the CAT Review Platform?

A: It's an AI-powered exam preparation platform designed specifically for Filipino Certified Accounting Technician (CAT) examinees. It uses Retrieval-Augmented Generation to provide personalized tutoring based on actual study materials.

Q: How does the AI tutor work?

A: The platform processes review materials into vector embeddings, then uses semantic search to find the most relevant content when a student asks a question. The AI synthesizes this into clear, exam-focused guidance.

Q: Is this platform only for CAT exams?

A: The platform was built with white-label architecture. The same core technology now powers cpareview.ph for CPA exam preparation, and can be expanded to other PRC-administered licensure exams.

Q: What technologies were used to build it?

A: The platform uses Next.js 15, PostgreSQL with pgvector for vector search, Anthropic's Claude API for AI tutoring, and Railway for cloud deployment.