Ivan Zepeda

Augmented Reality Experience for 1D Fans.

Project Management, Execution, Procurement, Customer Attention, Client Meetings, Content Creation, Navigational Map, Quality Assurance, Go Live.

Augmented Reality App and Kiosk.

Project Management, Execution, Procurement, Customer Attention, Client Meetings, Content Creation, Navigational Map, Quality Assurance, Go Live.

Augmented Reality App.

Project Management, Execution, Procurement, Customer Attention, Client Meetings, Content Creation, Navigational Map, Quality Assurance, Go Live.

Mobile App, Video and Content Creation, Digital Magazine Development.

Marketing Services for the Company. Marketing Plan, Execution, Web Development, Project Manager, Project Development, Customer Attention, Client Meetings, Content Creation, Quality Assurance.

Augmented Reality App.

Marketing Services for the Company. Marketing Plan, Execution, Web Development, Project Manager, Project Development, Customer Attention, Client Meetings, Content Creation, Quality Assurance.

Movie identification using image recognition

Project Overview:

"Kadabra" is an advanced multimodal movie identification system designed to recognize movies based on a single image or a brief text description. The system uses facial recognition technology to identify actors in a photo, cross-references them with a database of movies, and accurately matches the actors' ages and appearance to specific films. Users can also input a movie description via text or voice, and the system… Show more

Movie identification using image recognition

Project Overview:

"Kadabra" is an advanced multimodal movie identification system designed to recognize movies based on a single image or a brief text description. The system uses facial recognition technology to identify actors in a photo, cross-references them with a database of movies, and accurately matches the actors' ages and appearance to specific films. Users can also input a movie description via text or voice, and the system uses natural language processing (NLP) to provide the best possible movie match.

Kadabra is available through a Telegram bot, making it easily accessible for users to identify films on the go.

Key Features:

Facial Recognition: Detects actors from an image and compares them to a database using convolutional neural networks (CNNs) for accurate identification.
Movie Matching: Cross-references actors and movie data to identify the film, factoring in actors' age, appearance, and historical filmography.
Text & Voice Input: Uses NLP models like SpaCy and NLTK to process text or voice descriptions of the movie plot, generating probable matches based on semantic similarity.
Multi-Input Functionality: Users can either provide a photo of the actors or a written or spoken plot description to search for movies.
Accurate Filtering: Matches are narrowed down by comparing actor ages and timelines, improving the system’s precision for older films.
User-Friendly Output: Results are presented in a clear, detailed PDF format that includes movie information and visuals for easy identification.
Technologies Used:

Python
TensorFlow, Keras, PyTorch: For building facial recognition and machine learning models.
CNN: To perform facial recognition on actor images.
SpaCy, NLTK: For NLP-based text analysis to match movie descriptions with the database.
Telegram API: Chatbot - interface

This system provides a seamless way to identify movies based on various inputs, making movie discovery both fun and efficient. Show less

Led the development of Recipix, a multimodal system that combines computer vision and NLP to analyze images of fridge contents and generate personalized recipes. The system identifies ingredients using advanced computer vision models (ResNet-50, YOLO, Segment Anything Model) with an 81% classification accuracy across 88 food categories.

Key Highlights:

Built and optimized image classification models on a curated dataset of 41,000 images, improving recognition of diverse food… Show more

Led the development of Recipix, a multimodal system that combines computer vision and NLP to analyze images of fridge contents and generate personalized recipes. The system identifies ingredients using advanced computer vision models (ResNet-50, YOLO, Segment Anything Model) with an 81% classification accuracy across 88 food categories.

Key Highlights:

Built and optimized image classification models on a curated dataset of 41,000 images, improving recognition of diverse food items.
Applied image augmentation techniques and GPU acceleration (CUDA) to enhance model performance and training speed.
Developed an NLP-powered recipe recommendation engine using SpaCy, Word2Vec, and NMF, factoring in user preferences such as dietary restrictions and allergies.
Deployed the system with real-time user interaction through FastAPI and Telegram Bot API, allowing image and text-based input for recipe suggestions.
Integrated Docker and Celery with Redis to streamline processing pipelines, reducing image processing time from 5 minutes to 5 seconds.
Leveraged web scraping and backend systems (Django, PostgreSQL) to expand and manage the recipe database.
Technologies: Python, PyTorch, TensorFlow, OpenCV, FastAPI, Django, SpaCy, YOLO, Segment Anything Model, NMF, Docker, CUDA. Show less

Developed an NLP Medical Chatbot using NLTK, Spacy, and Whisper on Telegram.
It processed text/audio inputs,
Running different models detect semantic weight based on NER (entities), user info and symptoms,
calculating BMI, and identifying diseases by assessing the information against a database. Rigorous
preprocessing improved text comprehension. The chatbot generated diagnostic PDFs with user info,
symptoms, diseases, and treatments, showcasing my NLP, AI, and healthcare… Show more

Developed an NLP Medical Chatbot using NLTK, Spacy, and Whisper on Telegram.
It processed text/audio inputs,
Running different models detect semantic weight based on NER (entities), user info and symptoms,
calculating BMI, and identifying diseases by assessing the information against a database. Rigorous
preprocessing improved text comprehension. The chatbot generated diagnostic PDFs with user info,
symptoms, diseases, and treatments, showcasing my NLP, AI, and healthcare technology fusion expertise

https://github.com/ijzepeda-LC/CareNavi Show less