⭐ PRODUCT + PROBLEM WE SOLVE We built Udemy but for Medical GxP field - all from scratch as a full-scale EdTech platform for clinical research professionals, combining structured learning, team training, and advanced AI workflows in one product. The platform was designed to help CRA teams, auditors, coordinators, and sponsors work faster with complex GxP and ICH guidance while also making training more practical, scalable, and measurable. At the product level, the solution includes certified courses, corporate learning workflows, progress tracking, dashboards for organizations, and a flexible subscription model that combines training content with AI access. Instead of building just another LMS, we focused on a domain-specific platform where education and day-to-day professional decision support work together. 🤖 CUSTOM AI, IS THIS JUST ANOTHER BORING AI OR SOMETHING MORE? The main innovation was the custom AI layer we developed for the client. We implemented a two-tier AI system: AI Basic for fast regulatory and GCP-related questions AI Pro for deeper expert workflows powered by RAG and scenario-specific logic For the AI Pro layer, we built a custom RAG pipeline connected to the client’s proprietary knowledge base - more than 500 pages of unique source materials provided by the client. These documents were ingested, chunked, embedded, and structured so the assistant could retrieve relevant information from real PDFs and DOCX files instead of relying only on general model knowledge. This made the product much more accurate, much more defensible in regulated use cases, and far more valuable than a generic chatbot. The underlying platform spec explicitly describes the RAG setup, real-document ingestion, configurable models, editable system prompts, and persistent chat history. 👨⚕️ NOT JUST AI, BUT HUMAN-IN-THE-LOOP + AI We also developed a specialized AI copilot for RBQM, which guides users through a controlled multi-step workflow (which has to be confirmed by human ): Define CtQFs Generate QTLs Build RACT Review and improve outputs This workflow is especially important because it turns AI from a simple Q&A tool into a practical working assistant for real compliance and risk-management tasks. The system enforces dependencies between steps, requires a Study Profile at the start, marks assumptions with [ASSUMED], and references regulatory logic in the output. That makes it much closer to a domain copilot than a standard prompt wrapper. ❗IS THIS AI SAFE? WHAT ABOUT DATA LEAK? To support stricter enterprise and privacy requirements, we also implemented a secure anonymized AI layer on a dedicated local LLM environment. This added a protected path for sensitive workflows, where client data can be processed through an anonymization layer before model interaction. For regulated industries like pharma and clinical research, this architecture is a major advantage because it improves control over confidentiality, deployment boundaries, and trust in AI-assisted processes. 🎯 RESULT NOW: PRODUCT THAT ALREADY HAS 100+ USERS As a result, The QARP Academy became a vertical product at the intersection of EdTech, compliance software, and AI copilots. It does not just teach users theory - it helps them apply it in practice, automate expert workflows, and reduce work that would normally take teams days or weeks into a guided, AI-assisted process. The strongest demo scenario is the RBQM build flow, where the system can move from a raw study profile to a structured framework with detailed outputs, tables, and regulatory grounding.
Open case