Hello!
I'm John Ivan Diaz, a passionate Machine Learning Engineer who loves building intelligent systems
Hi! I’m John Ivan, a Professional Machine Learning Engineer and award-winning researcher who loves building ML systems. I’m passionate in artificial intelligence due to the profound impact it has done to us. I have strong knowledge in end-to-end ML lifecycle. I have hands-on experience defining use cases, developing ML models, and applying MLOps practices that promote Responsible AI. I have worked with global software engineering teams and have experience working at scale. I’m a fresh graduate, but the track record, Cum Laude, publications, awards, and leadership roles, shows I learn fast and committed to excellence. I’m eager to commence my professional journey with reputable organization dedicated to world class quality.
A deep learning model trained to classify CIFAR-100 images of reptiles into five categories: crocodile, dinosaur, lizard, snake, and turtle. It uses convolutional, pooling, and fully connected layers, along with techniques such as data augmentation, dropout, and early stopping. The model achieved decent accuracy despite the challenge of limited training images.
A multimodal deep learning model that predicts the next frame of plant growth by understanding two input modalities: growth patterns and environmental conditions. For instance, the model changes its prediction when the user adjusts pH level. Images of growth patterns pass through convolutional long short-term memory (ConvLSTM) layers, while environmental data passes through a series of layers such as time-distributed, reshape, and lambda. This produces two tensors with aligned temporal dimensions, which are concatenated to become a combined multimodal tensor. To generate the predicted frame, a 3D Convolution layer is employed.
In addition, a custom loss function is used to leverage the temporal dynamics and visual fidelity of Temporal Consistency and Mean Squared Error losses. The weighted sum of both is evaluated against each standalone loss quantitatively using metrics like Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index (SSIM), and Total Variation, and qualitatively through visual inspection. Results show that the custom loss delivers slightly better performance in preserving visual fidelity and temporal dynamics of slow-moving objects.
Two deep learning models trained on popular segmentation architectures, YOLOv8 and Detectron2, for segmenting the plant class. A dataset of plant images with diverse types and backgrounds was collected, and both models were trained using identical hyperparameters, then adjusted in parallel to compare performance, training speed, and memory consumption. Quantitative metrics such as average precision, loss, training time, RAM, and disk usage were used for evaluation. Results show that YOLOv8 trains faster and maintains consistent performance, while Detectron2 achieves higher accuracy when using larger hyperparameter settings.
A computer vision algorithm that reconstructs a 3D model from constrained 2D stereoscopic images captured from the front, left, back, and right of an object. Unlike traditional photogrammetry that requires many angles, this approach is designed to address limited views. It combines standalone computer vision techniques into a unified process, leveraging OpenCV and Open3D libraries. It is useful in applications such as plant phenotyping, where capturing multiple angles may be challenging yet still sufficient for constructing 3D models for visual inspection and analysis.
A simple desktop application that predicts the next path of a typhoon based on observed data points of its previous trajectory. Python and PyQt were used for full-stack development. Although the forecasting is limited to historical path data and does not account for other factors influencing a typhoon’s movement, the application is primarily designed to illustrate the concepts of interpolation and extrapolation.
- Python
- Java
- C
- SQL
- MATLAB
- TensorFlow
- Keras
- Scikit-Learn
- Computer Vison
- LLM
- BigQuery
- Apache Beam
- Dataflow
- Pub/Sub
- Artifact Registry
- GCP
- Vertex AI
- Kubernetes
- Docker
- Git/GitHub
- Collaborated with global software engineering teams across time zones working on a procurement platform.
- Delivered mission-critical KPI reports using SAP Ariba, giving stakeholders the visibility to act quickly, minimize financial risks, and maintain 24/7 operational continuity.
- Monitored automation using Control-M and flagged system failures in real time, accelerating team response and reducing resolution delays.
- Communicated system performance accurately, reinforcing fairness and guaranteeing dependable client transactions.
Presented my research titled “Innovative 3D Model Reconstruction from 2D Stereoscopic Plant Images with Investigation between YOLOv8 and Detectron2 AI Architectures for Plant Class Segmentation” as an oral presenter at The 10th International Conference on Next Generation Computing (ICNGC 2024), held at the Holy Angel University Auditorium, Angeles City, Philippines, in November 2024. The conference was organized by the Korean Institute of Next Generation Computing (KINGPC), a South Korea-based research organization dedicated to advancing research and development in emerging computing technologies.
The research solves a computer vision challenge of creating a 3D model of a plant from constrained stereo images for visualizing plant growth structures, that can be useful for plant phenotyping.
Awarded with Best Paper Award in Oral Session, recognized among 80 international research presenters.
Presented my research titled “Investigating Deep Learning and Computer Vision for Predicting and Simulating Plant Growth Structures: Laying the Groundwork for Digital Twins in Agriculture,” as an oral and poster presenter at The 9th University Research Conference (URC 2025), held at the University of San Carlos Fr. Michael Richartz Conference Center, Cebu City, Philippines, in April 2025.
The research lays groundwork in pursuing digital twin technology in agriculture - a digital replicas of crops in farms that mimic their behaviors and appearances from physical world to digital world, using artificial intelligence and deep learning. It aims to help farmers make informed farming decisions, and redefine farming practices to be more sustainable and efficient, promoting the United Nations Sustainable Development Goal (SDG) 2 (Zero Hunger), SDG 9 (Industry, Innovation, and Infrastructure), and SDG 12 (Responsible Consumption and Production), that overall benefits all of humanity and the planet.
URC is the University of San Carlos’ largest research event.
Awarded with Best Poster Presenter Award, recognized among more than 50 poster presenters from various departments and colleges of the university.
Proceedings of The 10th International Conference on Next Generation Computing.
HSST Excellence With Honors Graduate
Soonchunhyang University
Presented and worked my research in machine learning, deep learning, and computer vision together with international students from Soonchunhyang University (SCH), South Korea, demonstrating my ability to build relationships, exchange ideas, and collaborate amidst language barriers.
This activity was part of SCH’s Cultural Internship Program, where SCH students visit partner universities abroad, including the University of San Carlos (USC) of the Philippines, for research exposure, cultural exchange, and academic collaboration.
The collaboration took place for 10 days (July 2-12, 2024) at the University of San Carlos, where I was one of the few selected undergraduate researchers to collaborate with the South Korean students. My engagement helped strengthened the partnership between the two universities and fulfill the program’s vision.
Holy Spirit School of Tagbilaran
Served as Class President for eight consecutive years, from Grade 5 to Grade 12. Demonstrated leadership by helping classmates ignite teamwork in class activities, and served as liaison between classmates and faculty to facilitate communication between the two groups.
HSST Robotics
Served as an ambassador of HSST Robotics from 2019 to 2021, helping the school launch its inaugural robotics program through immersive hands-on experiences with various robotics activities and competitions across the country. Engaged with robotics experts and shared the knowledge gained back to the school to support the development of its new curriculum.
HSST Robotics is the official robotics team of Holy Spirit School of Tagbilaran (HSST), a private school in Bohol, Philippines. It aimed to introduce robotics as an avenue to enhance creativity and problem-solving skills among students.
HSST Media
Served as President of HSST Media from 2018 to 2021 that led in ensuring that documentations are high quality and are delivered punctually. Guided the team to recognize that our work’s sense of purpose is to communicate the school’s vision and mission.
HSST Media is the official photo and video documentations team of Holy Spirit School of Tagbilaran (HSST), a private school in Bohol, Philippines.
The Voice
Served as a Creative Director of The Voice from 2017 to 2021 that spearheads storytelling through design that has careful attention to detail, and collaborates with fellow editors and writers to ensure that the intended output is successfully delivered, all while ensuring it meets the publishing deadline.
The Voice is the official student publication of Holy Spirit School of Tagbilaran (HSST), a private school in Bohol, Philippines.
Ivan Studios
Built Ivan Studios as a personal creative outlet alongside my Computer Engineering studies. Worked with local and international clients (e.g., United Kingdom) by sharing my talents in photography, videography, and graphic design, and delivering high quality outputs, demonstrating my ability to balance study and work, manage time efficiently, and communicate effectively.
Awarded with Best Poster Presenter Award at The 9th University Research Conference (URC 2025). The award-winning research lays groundwork in pursuing digital twin technology in agriculture - a digital replicas of crops in farms that mimic their behaviors and appearances from physical world to digital world, using artificial intelligence and deep learning. It aims to help farmers make informed farming decisions, and redefine farming practices to be more sustainable and efficient, promoting the United Nations Sustainable Development Goal (SDG) 2 (Zero Hunger), SDG 9 (Industry, Innovation, and Infrastructure), and SDG 12 (Responsible Consumption and Production), that overall benefits all of humanity and the planet.
University Research Conference (URC) is the University of San Carlos’ largest research event.
The award was given at the University of San Carlos Fr. Michael Ritchartz Conference Center, Cebu City, Philippines, recognized among more than 50 poster presenters from various departments and colleges of the university.
Awarded with Best Paper Award in Oral Session at The 10th International Conference on Next Generation Computing (ICNGC 2024). The award-winning research solves a computer vision challenge of creating a 3D model of a plant from constrained stereo images for visualizing plant growth structures, that can be useful for plant phenotyping.
The conference was organized by the Korean Institute of Next Generation Computing (KINGPC), a South Korea-based research organization dedicated to advancing research and development in emerging computing technologies.
The award was given at the Holy Angel University auditorium, Angeles City, Philippines, recognized among 80 international research presenters.
Awarded with EduLearn Excellence in ICT Award by EduLearn Technologies Inc. for displaying commendable skills and leadership in various information and communications technology activities.
EduLearn Technologies Inc. is a Philippine-based education technology company specializing in robotics education and digital learning solutions known for organizing the annual ICT Robotics Competition for STEM students nationwide.
The award was given during the 2021 commencement rites at Holy Spirit School of Tagbilaran as a sole recipient among over 150 graduates.
Awarded with EduLearn Excellence in Robotics Award by EduLearn Technologies Inc. for displaying commendable skills and leadership in various robotics activities.
EduLearn Technologies Inc. is a Philippine-based education technology company specializing in robotics education and digital learning solutions known for organizing the annual ICT Robotics Competition for STEM students nationwide.
The award was given during the 2021 commencement rites at Holy Spirit School of Tagbilaran as a sole recipient among over 150 graduates.
Awarded with Best in Computer Award for displaying commendable skills and leadership in various computer-related activities that contributed positively to the school and local communities.
The award was given during the 2019 commencement rites at Holy Spirit School of Tagbilaran as a sole recipient among over 150 graduates.
Awarded with TechFactors Most Outstanding IT Student Award for displaying commendable skills and leadership in various information technology activities.
TechFactors Inc. is a leading e-learning company in the Philippines, collaborating over 400 private schools and impacting over 300,000 students nationwide, dedicated towards enhancing education through innovative digital solutions.
The award was given at Holy Spirit School of Tagbilaran as a sole recipient among over 150 students of the batch.
Awarded with Best in Computer Award for displaying commendable skills and leadership in various computer-related activities that contributed positively to the school and local communities.
The award was given during the 2015 commencement rites at Holy Spirit School of Tagbilaran as a sole recipient among over 150 graduates.
Awarded with Campus Journalism Award for displaying commendable skills and leadership in journalism, such as writing, photojournalism, and publication design, that contributed positively to the school’s success in representing and championing in various inter-school competitions.
The award was given during the 2021 commencement rites at Holy Spirit School of Tagbilaran as one of only three recipients among over 150 graduates.
Awarded with Saint Joseph Freinademetz Award, the highest award given upon a graduate of Holy Spirit School of Tagbilaran, for embodying holistic excellence exemplified by the life and ideals of Saint Joseph Freinademetz. This honor reflects a balanced integration of intellect, heart, and soul, that aligns with the school’s mission to develop well-rounded individuals dedicated to making a positive impact in the world.
The award was given during the 2021 commencement rites at Holy Spirit School of Tagbilaran as a sole recipient among over 150 graduates.
Awarded with Saint Joseph Freinademetz Award, the highest award given upon a graduate of Holy Spirit School of Tagbilaran, for embodying holistic excellence exemplified by the life and ideals of Saint Joseph Freinademetz. This honor reflects a balanced integration of intellect, heart, and soul, that aligns with the school’s mission to develop well-rounded individuals dedicated to making a positive impact in the world.
The award was given during the 2019 commencement rites at Holy Spirit School of Tagbilaran as a sole recipient among over 150 graduates.
Awarded with Saint Arnold Janssen Award, the highest award given upon a graduate of Holy Spirit School of Tagbilaran, for embodying holistic excellence exemplified by the life and ideals of Saint Arnold Janssen. This honor reflects a balanced integration of intellect, heart, and soul, that aligns with the school’s mission to develop well-rounded individuals dedicated to making a positive impact in the world.
The award was given during the 2015 commencement rites at Holy Spirit School of Tagbilaran as a sole recipient among over 150 graduates.