Overall
IOE Design AI is a product design project for an AI service hub platform, focusing on six core technologies: Text-to-Speech, eKYC, OCR, Virtual Human, Speech-to-Text, and Scoring. Based on research into the real-world needs of students, teachers, and the IOE operations team, I designed the end-to-end architecture: from a service-first information architecture and an "Experience" card allowing users to try AI immediately, to detailed workflows across Web, Mobile, and CMS platforms. The goal wasn't just to "add AI to the app"—but to transform AI into a clear, trustworthy, and user-friendly service within the digital education ecosystem.

Design System
~150–200+ screens across 5 pages · 6 AI services · Dark mode design system with blue-purple gradient, card-based layout, 3D isometric icons · Component library reused from hub card to CMS form builder · End-to-end design: Mobile + Web + Admin CMS.

When designing IOE Design AI, I didn't start with technology—but with the question: what difficulties are users facing, and what AI service can solve them? Below is a comprehensive matrix in text form, mapping each pain point to a specific product solution.

Manual grading, slow and overwhelming, is the biggest pain point for teachers and the operations team. Hundreds of essays and speeches each week lead to delayed feedback, and students lose motivation to correct their mistakes. Scoring directly addresses this problem: AI automatically grades assignments immediately after submission, analyzing details by skill—Listening, Reading, Grammar, Vocabulary—instead of just returning a total number. Teachers only need to review edge cases, not grade each assignment individually.

Practicing speaking without feedback is a particular pain point for students. At home, without a teacher or peer to practice with, they speak alone, unsure of their pronunciation or their current level. Virtual Human provides a virtual character for direct interaction: practice speaking, receive feedback and real-time scoring, creating the feeling of having a tutor even when studying alone. Speech to Text further enhances this: recording your voice and converting it to text allows students to compare and correct pronunciation errors immediately.

Not knowing which skills are weak leads students to blindly practice for exams – doing practice tests continuously without improving in the right areas. Scoring not only provides scores but also detailed reports on strengths and areas for improvement for each skill, helping students adjust their learning path with goals instead of cramming indiscriminately.

Cheating in online exams is a major concern for organizers when thousands of candidates take the exam simultaneously – insufficient supervisors and difficulty verifying who is actually taking the test. eKYC addresses this problem by verifying identity through documents and facial recognition before and during the exam – transparent, automated, and reducing the need for manual supervision.

Data entry for exam questions is time-consuming and prone to errors, a daily burden for teachers. Scanning paper and manually typing questions into the system is slow, prone to duplication, and error-prone. OCR allows uploading exam questions in image or PDF format, with AI extracting text, questions, and answers automatically and importing them directly into the CMS—where teachers select question types, add tables, and download the exam without having to retype everything from scratch.

The slow and expensive production of educational content is a barrier when IOE wants to expand its curriculum. Creating video lectures with presenters requires a production team, studio, and a long time. Text-to-Speech shortens the entire process: script input, AI creates videos with virtual presenters—producing large-scale educational content without live filming.

The perceived remoteness and inaccessibility of AI is a common psychological pain point—students, teachers, and even developers have heard about AI but don't know where to start, fearing the technology's complexity. The AI ​​service hub design features six cards—each representing a service, described in a single line, and a “Try It” button—transforming AI from an abstract concept into a usable tool, eliminating the need for lengthy documentation.

Dispersed exam data, difficult to aggregate, wastes organizers' time in comparing results. Speeches, written assignments, and uploaded files are scattered across multiple locations, making reporting and retrieval difficult. OCR and Speech-to-Text standardize all input into centralized digital data; Virtual Human and Scoring store all interaction and scoring results in a single system, providing real-time reporting without manual aggregation.

Integrating AI into a proprietary system is a major challenge for schools and partners wanting to use IOE technology but lacking a clear API and a pre-go-live testing workflow. The SDK download page and developer portal provide a complete onboarding process: view demos of each AI service → read documentation → download SDK → test in the sandbox—before integrating into their systems.

In short, the six AI services in IOE Design AI don't exist in isolation—they fit together like pieces of a puzzle: Scoring and Virtual Human serve students; OCR and CMS free up teachers; eKYC protects exam integrity; Text-to-Speech expands content production capabilities; and SDK connects the entire ecosystem externally. End-to-end product design here means ensuring each piece is visible, tryable, and trustworthy—right from the first card the user sees.