Giuseppe Edgardo G. Catambing
Quezon City
Summary
To create and innovate things in this world to make it a greater place, using my knowledge, physics and electronics.
Overview
6
6
years of professional experience
Work History
Lead Programmer / AI Engineer/ Data Scientist
HOKEI
Subic, Zambales
04.2024 - 06.2024
- Duties and Responsibilities:
- AI computer vision
- AI
- Convolutional Neural Networks (CNN)
- AI Model Architectures
- Python
- Open CV, Pytorch, Tensorflow, Sklearn, Mediapipe
- Face shape recognition, Gender recognition, Age recognition, Skin tone recognition
- Sklearn, Tensorflow, Pytorch
- Linear Algebra, Calculus, Probability and Statics
- Face age prediction, gender recognition, Face shape recognition, and Skin tone recognition procedure and methodology
- Identifying features and targets
- Collection of relevant datasets from the internet and from the real actual.
- Annotating the data collected
- Data preprocessing and data wrangling
- Feature engineering, filtering and de-noising
- Split dataset to train and validation 80/20 or 70/30, or cross validation
- Create an AI model architecture or retrain a premade AI model architecture like Facenet, VGG
- Face age prediction
- Since it’s a time series and linear, we set Linear activation function for the last activation layer
- Since it’s classification, we use mean square error for loss/cost function for training
- Adjust the hyperparameters for optimization function
- After training is inference where you will evaluate the model
- Evaluate metrics mean absolute error, accuracy, precision, recall, specificity, f1 score, etc
- Evaluate with confusion matrix
- Gender Recognition
- Since it’s a binary classification, we set Sigmoid activation function for the last activation layer
- Since it’s classification, we use categorical crossentropy for loss/cost function for training
- Adjust the hyperparameters for optimization function
- After training is inference where you will evaluate the model
- Evaluate accuracy, precision, recall, specificity, f1 score, etc
- Evaluate with confusion matrix
- Face shape recognition
- To get more correct features, we use mediapipe to filter out unwanted features or data
- Since it’s a multiple classification, we set softmax activation function for the last activation layer
- Since it’s classification, we use categorical crossentropy for loss/cost function for training
- Adjust the hyperparameters for optimization function
- After training is inference where you will evaluate the model
- Evaluate accuracy, precision, recall, specificity, f1 score, etc
- Evaluate with confusion matrix
- Skin Tone
- To get more correct features, we use mediapipe to filter out unwanted features and selecting a certain area
- Normalizing the lighting
- Using Ycrcb color space
- Using lTA for skin tone
- Since it’s a multiple classification, we set softmax activation function for the last activation layer
- Since it’s classification, we use categorical crossentropy for loss/cost function for training
- Adjust the hyperparameters for optimization function
- After training is inference where you will evaluate the model
- Evaluate accuracy, precision, recall, specificity, f1 score, etc
- Evaluate with confusion matrix
Data Scientist
MegaXcess (IT Solutions)
Pasig
10.2022 - 04.2024
- Duties and Responsibilities:
- Data Analytics process automation and calculations (automated)
- Data Wrangling and pulling of data automated
- Automated web scrapping
- Train models for images for authentication
- AI models
Project Technical Specialist I
DOST- Advance Science Technology Institute (ASTI)
UP Diliman, Diliman, Quezon City
01.2022 - 09.2022
- Duties and Responsibilities:
- Image Processing - Sklearn
- ROS 2 - Tensorflow
- YOLO - Pytorch
- AI models - Linear Algebra, Calculus, Probability, and Statics
- Object tracing and detection procedure and methodology
- Identifying features and targets
- Collection of relevant datasets from the internet and from the real actual.
- Annotating the data collected
- Data preprocessing and data wrangling
- Feature engineering, filtering and de-noising
- Split dataset to train and validation 80/20 or 70/30, or cross validation
- Create an AI model architecture or retrain a premade AI model architecture like YOLO
- Since it’s a multiple classification, we set Softmax activation function for the last activation layer
- Since it’s classification, we use categorical crossentropy for loss/cost function for training
- Adjust the hyperparameters for optimization function
- After training is inference where you will evaluate the model
- Evaluate metrics accuracy, precision, recall, specificity, f1 score, etc
- Evaluate with confusion matrix
Research Assistant
UST – RCNAS
UST Espana, Manila
02.2021 - 11.2021
- Duties and Responsibilities:
- Fall Detection Project
- Arduino nano 33 IOT
- Fall detection using Gyroscope, Accelerometer, ECG, and PPG.
- Data Analysis of Gyroscope, Accelerometer, ECG, and PPG using Python for fall detection
- Wireless data transfer (IOT) from Arduino nano 33 IOT (Webserver) to Blynk cloud application
- Development of a Wireless-based Armband physiological monitor for simultaneous Ballistocardiogram, Electrocardiogram and Photoplethysmogram signal acquisition
Researcher/Engineer/Scientist
Xcellcomms Enterprises
Novaliches, Quezon City
10.2019 - 01.2021
- Duties and Responsibilities:
- Solar powered alcohol dispenser
- Sensorized Alcohol Dispenser
- Hybrid Solar Panel (Under Development)
- Face Recognition and Fingerprint security door lock that sends SMS to the owner (used for factories or companies)
Technical Engineer
Makerlab Electronics
Novaliches, Quezon City
11.2018 - 02.2019
- Duties and Responsibilities:
- Made projects for some electronic enthusiasts.
- Made thesis projects for college students
- Doing thesis projects for students and electronic enthusiasts about Rpi and Arduino that uses sensors, electronic modules and electronic components.
- Learned more about the products of some electronic components and modules.
Education
Masters Degree - Electronics Engineering
University of Santo Tomas
09.2025
Bachelor of Science - Applied Physics, Major in Instrumentation
University of Santo Tomas
06.2019
Skills
- Advanced Skills in Electronics
- Skilled at Circuit Analysis
- Skilled at using Verilog for designing and verifying digital circuits
- Possesses impressive qualitative and quantitative analysis skills
- Skilled at Physics Analysis
- Equipped with Technical Skills
- Enhancing skills at 3D designing and 3D printing
- Skilled at using various programming languages such as:
- C/C
- Python
- Arduino
- MatLab
- Displays proficiency in utilization of fundamental machine learning skills
- Selected machine learning libraries for different tasks
- Designed new models and algorithms for machine learning projects
- Implemented machine learning experiments that lead to the development of new algorithms
- Tesda NC II Automotive Servicing
Timeline
Lead Programmer / AI Engineer/ Data Scientist
HOKEI
04.2024 - 06.2024
Data Scientist
MegaXcess (IT Solutions)
10.2022 - 04.2024
Project Technical Specialist I
DOST- Advance Science Technology Institute (ASTI)
01.2022 - 09.2022
Research Assistant
UST – RCNAS
02.2021 - 11.2021
Researcher/Engineer/Scientist
Xcellcomms Enterprises
10.2019 - 01.2021
Technical Engineer
Makerlab Electronics
11.2018 - 02.2019
Bachelor of Science - Applied Physics, Major in Instrumentation
University of Santo Tomas
Masters Degree - Electronics Engineering
University of Santo Tomas