One system. One language. One tap to anything.

The redesign unified all ERP modules under a single mobile-first interface built around the student's daily flow — login once, see everything relevant on the dashboard, drill into any module in one tap. A shared component library (cards, list rows, status chips) creates visual consistency across academics, accounts, and results.

The process began with a full interface audit of the existing system — cataloguing every screen, navigation pattern, and interaction. This was followed by an information architecture review to understand how data was structured and surfaced across modules. Workflow mapping helped identify where students encountered the most friction, and a component inventory revealed the extent of visual inconsistency between sections.

The redesigned ERP consolidates six fragmented modules into one consistent mobile experience. A shared design token set — a single blue primary, neutral grays, and one type scale — ensures every screen feels like part of the same product. Navigation cost dropped from 4–6 taps to a maximum of 2.

Navigation depth was reduced by restructuring the information architecture around student tasks rather than administrative modules. The most-accessed information — timetable, attendance, fees, and results — is now surfaced at the dashboard level, eliminating the intermediate module selection step that previously added friction to every session.

Discoverability improved significantly by replacing module-level menus with a dashboard that proactively surfaces relevant content. Students no longer need to know where a piece of information lives in the system — the interface presents it in context, reducing the time spent searching and the number of navigations required per session.

The unified visual language means students encounter the same card structure, status chips, and spacing conventions across every module. This consistency reduces the learning curve when moving between sections and creates a coherent experience that feels like a single, considered product rather than a collection of disconnected tools.

Six phases. One unified prescription pipeline.

PROSCRIPT was built in structured phases, each layer adding intelligence on top of the last. OCR converts raw images into text; NLP normalises that text into medical entities; retrieval and rule engines validate safety; and a large language model provides contextual assistance — all within a single API-backed workflow.

The decision to separate each processing stage into an independent service made the system easier to test and iterate. Each phase could be validated in isolation before integrating into the full pipeline, reducing the surface area for bugs in a domain where accuracy is critical.

PROSCRIPT demonstrates how OCR, machine learning, retrieval systems, and large language models can work together to improve prescription understanding and medication safety.

The platform successfully transforms fragmented prescription information into a structured workflow that supports interpretation, validation, and contextual assistance — replacing a manual, error-prone process with an automated pipeline that surfaces the right information at each step.

By separating concerns across six discrete processing phases, the system remains extensible. Each layer can be improved or swapped independently as better models or data sources become available, without requiring changes to the full stack.