ANDRZEJ MAREK
Back to selected work
Tech & Team LeadVERIFIED PROJECT

Elite Medical Prep — AI Tutor Platform

A production Socratic AI Tutor used by hundreds of medical students across more than a thousand lessons.

01 / Context

Delivered with Widelab for Elite Medical Prep, the web application reduces reliance on expensive one-to-one physician tutoring. A student first answers an open medical question and then continues in a chat with an AI Tutor. Rather than immediately revealing the answer, the tutor evaluates the response, asks follow-up questions and guides the student through a Socratic learning flow. Tutors and administrators prepare lessons and source material, while the platform produces knowledge-gap reports that tutors can review when recommending the next learning steps. As Tech and Team Lead in a four-person team, Andrzej owned client discovery, architecture and hands-on full-stack delivery through production deployment.

02 / Scope and decisions
  • Led a four-person delivery team while remaining hands-on across React, Node.js, TypeScript, Python and AWS.
  • Built the open-question and conversational lesson flow around Socratic teaching rather than direct answer generation.
  • Used prompts, a state machine, lesson rules and dedicated classification agents to evaluate responses and control lesson progression.
  • Implemented RAG and vector retrieval over books, medical materials and content prepared by tutors and administrators.
  • Used PostgreSQL as both the product database and the store supporting vector retrieval.
  • Separated Node.js product APIs and integrations from the Python AI, agent, RAG and classification layer.
  • Integrated OpenAI, Anthropic, Google Gemini and Amazon Bedrock model providers.
  • Used WebSockets for interactive application communication.
  • Deployed the production platform with AWS ECS, RDS, S3, CloudFront, SQS and Bedrock.
  • Delivered student, tutor and administrator experiences, a lesson editor, reporting, analytics and user management.
03 / Outcomes
  • Built and launched the platform from requirements through production deployment.
  • Supported hundreds of medical students and more than a thousand lessons.
  • Enabled tutors to review knowledge-gap reports and recommend subsequent learning steps.
04 / Technology

React · TypeScript · Node.js · Python · PostgreSQL · WebSocket · RAG · Vector Search · OpenAI · Anthropic · Google Gemini · Amazon Bedrock · AWS ECS · Amazon RDS · Amazon S3 · Amazon CloudFront · Amazon SQS

AMI / PROJECT CONTEXT

Ask about the decisions, ownership and delivery.

Ask Ami about this projectVisit project