I am Angelica
Data Scientist & Artificial Intelligence Engineer
My expertise in blending mathematics, IT, and business insights enables me to extract meaningful value from data, paving the way for innovative solutions and advancements in AI-driven analytics. By using these powerful tools, I aim to drive impactful decision-making in industry. Lastly, I am a FLASK lover.
My Collaboration Projects
As a full-stack developer, I've partnered with various events and companies to create comprehensive web solutions. These collaborations showcase my expertise in developing end-to-end applications using Flask and Python, encompassing backend, frontend, database, system architecture, and AI integration.
AI Face Swap
Gudang Garam International
Forestra'24
Attendance System
Schoters by Ruangguru
Study and Work Abroad Festival
Glasses Virtual Try On
Eyesoul Kawan Lama
Living World Alam Sutera
AI Mascot Chatbot
AMO Nabati
Interactive AI Assistant
Data Sciences Projects
Exploring Biological Age Models Prediction with Clinical
Biomarkers using Support Vector Regression and
Klemera-and-Doubal Method
Prediction
Statistical Method
Klemera-and-Doubal
Machine Learning
Support Vector Regression
SciPy
Scikit-Learn
Regression
Used Support Vector Regression and Klemera-and-Doubal Method for biological age modeling from clinical biomarkers. Conducted
multicollinearity and feature selection analyses, evaluating model performance with RMSE and R-squared metrics. Supported by
PUTI Q1 Research Grant, Universitas Indonesia. This work received support from the PUTI Q1 Research Grant 2023 provided by
Universitas Indonesia.
Detecting Retinopathy of Prematurity Disease in Premature Baby
Based on Fundus Image Data with CNN Model using VGG19's
Architecture
Image Classification
Deep Learning
Convolutional Neural Network
VGG-19 Architecture
Keras
TensorFlow
OpenCV
A Convolutional Neural Network, employing the VGG-19 architecture, is applied to detect Retinopathy of Prematurity (ROP) in premature infants, with emphasis on assessing the impact of data augmentation on model performance. The study involves comprehensive data preparation, including augmentation, labeling, and preprocessing of retinal images, followed by model training and evaluation to determine the most effective approach for ROP detection.
Credit Submission Detectors using Logistic Regression, Support Vector Machines, and Decision Tree
Supervised Learning
Pearson Correlation
Logistic Regression
Support Vector Machines
Decision Tree
Scikit-Learn
This project evaluates and compares the performance of three machine learning models - Decision Tree, Logistic Regression, and Support Vector Machine - for credit scoring in lending businesses, with a focus on predicting loan repayment. The Decision Tree model emerged as the best performer, achieving an F1-score of 94%, and the study concludes by highlighting key factors for determining credit eligibility, including annual income, loan class, interest rate, and default history.
Analysis of Gene Expression Detection in Classifying Types of Cancer: Acute Myeloid Leukemia (AML) and Acute Lymphoblastic Leukemia (ALL) using PCA and Support Vector Machine (SVM)
GEOparse
Scipy
Principal Component Analysis
Support Vector Machines
This project focuses on analyzing DNA methylation profiles in Acute Myeloid Leukemia (AML) and Acute Lymphoblastic Leukemia (ALL) using datasets from genome-wide studies. The methodology involves combining and preprocessing the data, applying dimensionality reduction through Principal Component Analysis, and then employing a Support Vector Machines Classifier for machine learning, with the model's performance evaluated using ROC AUC.
Mushroom Characteristic Classification using Artificial Neural Network (ANN) with Multilayer Perceptron (MLP) Architecture
Scikit-Learn
Keras
Multilayer Perceptron
Artificial Neural Network
A deep neural network is developed to classify mushroom characteristics, employing hyperparameter tuning to optimize the model's performance in identifying potentially toxic mushrooms. The project encompasses comprehensive data analysis, preprocessing, and feature selection, culminating in a highly accurate model with 99% accuracy and an F1-score of 1.00, while also pinpointing key features strongly associated with mushroom toxicity.
Exploring Biological Age Models Prediction with Clinical Biomarkers using Support Vector Regression and Klemera-and-Doubal Method
Used Support Vector Regression and Klemera-and-Doubal Method for biological age modeling from clinical biomarkers. Conducted multicollinearity and feature selection analyses, evaluating model performance with RMSE and R-squared metrics. Supported by PUTI Q1 Research Grant, Universitas Indonesia. This work received support from the PUTI Q1 Research Grant 2023 provided by Universitas Indonesia.
Detecting Retinopathy of Prematurity Disease in Premature Baby Based on Fundus Image Data with CNN Model using VGG19's Architecture
A Convolutional Neural Network, employing the VGG-19 architecture, is applied to detect Retinopathy of Prematurity (ROP) in premature infants, with emphasis on assessing the impact of data augmentation on model performance. The study involves comprehensive data preparation, including augmentation, labeling, and preprocessing of retinal images, followed by model training and evaluation to determine the most effective approach for ROP detection.
Credit Submission Detectors using Logistic Regression, Support Vector Machines, and Decision Tree
This project evaluates and compares the performance of three machine learning models - Decision Tree, Logistic Regression, and Support Vector Machine - for credit scoring in lending businesses, with a focus on predicting loan repayment. The Decision Tree model emerged as the best performer, achieving an F1-score of 94%, and the study concludes by highlighting key factors for determining credit eligibility, including annual income, loan class, interest rate, and default history.
Analysis of Gene Expression Detection in Classifying Types of Cancer: Acute Myeloid Leukemia (AML) and Acute Lymphoblastic Leukemia (ALL) using PCA and Support Vector Machine (SVM)
This project focuses on analyzing DNA methylation profiles in Acute Myeloid Leukemia (AML) and Acute Lymphoblastic Leukemia (ALL) using datasets from genome-wide studies. The methodology involves combining and preprocessing the data, applying dimensionality reduction through Principal Component Analysis, and then employing a Support Vector Machines Classifier for machine learning, with the model's performance evaluated using ROC AUC.
Mushroom Characteristic Classification using Artificial Neural Network (ANN) with Multilayer Perceptron (MLP) Architecture
A deep neural network is developed to classify mushroom characteristics, employing hyperparameter tuning to optimize the model's performance in identifying potentially toxic mushrooms. The project encompasses comprehensive data analysis, preprocessing, and feature selection, culminating in a highly accurate model with 99% accuracy and an F1-score of 1.00, while also pinpointing key features strongly associated with mushroom toxicity.
Artificial Intelligence Projects
Surveillance Tracking using YOLOv8, Counting using Supervision, and Recognizing using MTCNN and FaceNet
Computer Vision
Supervision
YOLOv8
MTCNN
FaceNet
Face Recognition
Face Identification
This project implements a comprehensive surveillance system using advanced computer vision techniques, combining YOLOv8 for object detection and tracking, Supervision for object counting, and MTCNN with FaceNet for face recognition and identification. The system addresses the challenges of Multiple Object Tracking (MOT) in complex scenarios, offering robust solutions for object localization, identity maintenance, and trajectory tracking across video frames, while also incorporating sophisticated face recognition capabilities to enhance surveillance and identification tasks.
Intelligent Document Processing (IDP) for Banking Document: Bank Statement, Financial Statement, and Trade Finance
Intelligent Document Processing
Optical Character Recognition
Natural Language Processing
PaddleOCR
Tesseract OCR
EasyOCR
Large Language Models
Generative Pre-trained Transformer
YOLOv8
Intelligent Document Processing (IDP) leverages advanced technologies like OCR, NLP, and machine learning to automate the extraction, analysis, and understanding of information from complex banking documents such as bank statements, financial statements, and trade finance records. This project employs a combination of state-of-the-art tools including PaddleOCR, Tesseract, EasyOCR, large language models, GPT, and YOLOv8 to create a robust system capable of efficiently processing and interpreting diverse financial documents, potentially revolutionizing document handling in the banking sector.
Biometric Access Control with Real-Time Face Recognition and Face Comparison using Landmark Facial Embeddings
Computer Vision
Face Recognition
Face Comparison
MTCNN
FaceNet
Dlib
Facial Embedding
Real-time Verification
Biometric Access Control
FacePass is an innovative web application that combines advanced AI technologies like MTCNN, FaceNet, and computer vision to provide secure and efficient event access control through real-time face recognition. The system streamlines the event check-in process by integrating QR code scanning with facial comparison, ensuring that only registered participants gain entry while effectively preventing unauthorized access.
NLP-Driven Task Recommendation Engine for Event Planning using TF-IDF
NLTK
Tokenization
Lemmatization
TF-IDF Vectorization
Jaccard Similarity
Cosine Similarity
This project implements a sophisticated Natural Language Processing (NLP) system that generates tailored task recommendations for event planners based on company profiles and event objectives. Leveraging advanced text processing techniques including tokenization, lemmatization, TF-IDF vectorization, and Jaccard similarity calculations, the engine analyzes textual inputs to produce highly relevant and context-specific task suggestions, enhancing the efficiency and effectiveness of event planning processes.
Poses Comparison between Two Photos using Movenet Multipose Lightning
Movenet
Multipose Lightning
TensorFlow
OpenCV
OpenPose
PoseNet
Keypoint Detection
This project implements a computer vision system for pose estimation and comparison using MoveNet MultiPose Lightning, integrated with TensorFlow and OpenCV. The system processes images from both event organizers and participants, extracts 17 key body landmarks, and calculates a detailed similarity score based on keypoint positions, enabling precise pose matching for applications like virtual event participation or fitness instruction verification.
Facial Analysis System for Eyewear Fitting using MediaPipe and Inception V3
Facial Analysis
MediaPipe Face Landmarker
Inception V3 Architecture
3D Facial Landmarks
Face Shape Classification
Gender Classification
Geometric Face Measurements
This facial analysis system employs MediaPipe for precise 3D facial landmark detection and geometric calculations, combined with Inception V3 architecture for face shape and gender classification. The system provides comprehensive facial measurements and shape analysis, enabling personalized eyewear recommendations and virtual try-on experiences, while also offering capabilities for facial expression recognition and avatar creation.
Surveillance Tracking using YOLOv8, Counting using Supervision, and Recognizing using MTCNN and FaceNet
This project implements a comprehensive surveillance system using advanced computer vision techniques, combining YOLOv8 for object detection and tracking, Supervision for object counting, and MTCNN with FaceNet for face recognition and identification. The system addresses the challenges of Multiple Object Tracking (MOT) in complex scenarios, offering robust solutions for object localization, identity maintenance, and trajectory tracking across video frames, while also incorporating sophisticated face recognition capabilities to enhance surveillance and identification tasks.
Intelligent Document Processing (IDP) for Banking Document: Bank Statement, Financial Statement, and Trade Finance
Intelligent Document Processing (IDP) leverages advanced technologies like OCR, NLP, and machine learning to automate the extraction, analysis, and understanding of information from complex banking documents such as bank statements, financial statements, and trade finance records. This project employs a combination of state-of-the-art tools including PaddleOCR, Tesseract, EasyOCR, large language models, GPT, and YOLOv8 to create a robust system capable of efficiently processing and interpreting diverse financial documents, potentially revolutionizing document handling in the banking sector.
Biometric Access Control with Real-Time Face Recognition and Face Comparison using Landmark Facial Embeddings
FacePass is an innovative web application that combines advanced AI technologies like MTCNN, FaceNet, and computer vision to provide secure and efficient event access control through real-time face recognition. The system streamlines the event check-in process by integrating QR code scanning with facial comparison, ensuring that only registered participants gain entry while effectively preventing unauthorized access.
NLP-Driven Task Recommendation Engine for Event Planning using TF-IDF
This project implements a sophisticated Natural Language Processing (NLP) system that generates tailored task recommendations for event planners based on company profiles and event objectives. Leveraging advanced text processing techniques including tokenization, lemmatization, TF-IDF vectorization, and Jaccard similarity calculations, the engine analyzes textual inputs to produce highly relevant and context-specific task suggestions, enhancing the efficiency and effectiveness of event planning processes.
Poses Comparison between Two Photos using Movenet Multipose Lightning
This project implements a computer vision system for pose estimation and comparison using MoveNet MultiPose Lightning, integrated with TensorFlow and OpenCV. The system processes images from both event organizers and participants, extracts 17 key body landmarks, and calculates a detailed similarity score based on keypoint positions, enabling precise pose matching for applications like virtual event participation or fitness instruction verification.
Facial Analysis System for Eyewear Fitting using MediaPipe and Inception V3
This facial analysis system employs MediaPipe for precise 3D facial landmark detection and geometric calculations, combined with Inception V3 architecture for face shape and gender classification. The system provides comprehensive facial measurements and shape analysis, enabling personalized eyewear recommendations and virtual try-on experiences, while also offering capabilities for facial expression recognition and avatar creation.
