DocMe Assistant
Chatbot to classify intents and manage medical appointments in Colombia
- Client
- Academic project · DocMe
- Role
- Backend Developer · Machine Learning
- Year
- 2024
Summary — I designed the architecture of a medical-appointment chatbot: TensorFlow/Keras model training, a Flask backend with SQLAlchemy and MySQL, and natural language processing with NLTK.
DocMe is a conversational assistant designed to ease overloaded medical hotlines: it classifies user intents, schedules appointments, and provides basic emotional support.
NLP and neural networks
I trained neural networks with TensorFlow and Keras to classify intents. Combined with NLTK and PySpellChecker, the bot interprets greetings, symptoms, emotional-support requests, and appointment requests, including spell correction.
- Models
- TensorFlow · Keras
- NLP
- NLTK · PySpellChecker
What I implemented
- End-to-end chatbot architecture, from the model to the API.
- Intent classification model training.
- Flask backend with SQLAlchemy and MySQL for users and appointments.
- Production server with Waitress (WSGI).
- NLP processing, spell correction, and response generation.