Medical Records Digitization Pipeline

Problem: Paper-based records block access to patients' long-term medical history at point of care.

Solution: LLM-powered OCR system for automated clinical data extraction.

Impact: Improves patient safety and treatment accuracy through timely access to historical clinical data.

Watch Demo

AI Symptom Checker & CDSS

Problem: Many patients face difficulty articulating their symptoms and identifying the right specialist.

Solution: A symptom checker that suggests relevant symptoms, asks targeted follow-up questions, and provides provisional diagnoses with specialist recommendations.

Impact: 81% diagnostic accuracy validated against physician assessments across 185 simulated patient cases.

Watch Demo

Medical History-Driven Dietary Recommendation System

Problem: Gap between visual food recognition technology and personalized medical dietary counseling for chronic disease management.

Solution: Integrated computer vision-based food classification with medical knowledge systems for condition-specific dietary filtering and recommendations.

Impact: Enables real-time food guidance, improving patient adherence to dietary restrictions for chronic conditions.

Watch Demo

Medical Assistant Chatbot

Problem: Users need to describe symptoms and access care without navigating complex interfaces.

Solution: Began as a RASA-based intent classifier for mental vs. physical health triage; evolved into an LLM-powered assistant serving as AmarDoctor's primary entry point for symptom assessment, appointment booking, and care navigation.

GitHub

Bengali OCR Using Deep Learning

Developed a novel word image dataset and trained deep neural networks (CNN, RNN, LSTM) to digitize printed Bengali literature.

Dataset

SaaS Client Management System

Built a JWT-authenticated platform with PostgreSQL and AWS for managing secure healthcare API access and subscription tiers.