I'm Juan Sebastian Estupiñan, I'm a Python developer with 2+ years of experience. Specialized in Machine Learning and Deep Learning. I'm currently working with the following technologies: Python, Numpy, Pandas, Scikit-Learn, Keras, Pytorch, Tensorflow. I'm a Software Engineer from Universidad Industrial de Santander (UIS). I worked in the academic environment in the High Dimensional Signal Processing (HDSP) research group, focusing on topics like Phase Retrieval, Medical Imaging, Binary Neural Networks and Image Processing.
I consider myself a good listener, team worker, patient and determined person. I'm interested in the potential of AI.
I have advanced English level, basic French level and I'm open to new professional challenges that allow me to improve my skills mainly in Machine Learning and Data Science areas.
You can contact me by email at juansec13@gmail.com
• Conducted data analysis and visualization using SQL, Pandas, Matplotlib, and Seaborn, facilitating data-driven decisions and insights.
• Orchestrated ETL pipeline creation for 2016-2018 revenue data analysis, improving data handling.
• Applied descriptive and inferential statistics in organized data to complete data-driven tasks.
• Implemented and applied Machine Learning algorithm like Random Forest and LightGBM to analyse and predict Credit Default Risk.
• Tested and completed a Machine Learning API based on a microservice architectures, using Redis, Flask, Docker and Tensorflow for animal classification.
• Achieved 82% accuracy in classifying cars with 25 classes using Convolutional Neural Networks and Transfer Learning (LeNET, ResNet50, EfficientNetB7), demonstrating strong proficiency in image classification tasks.
● Achieved 80% F1-Score in cross-validation tests with a Deep Learning Framework to assist in the treatment of chronic wounds from Hansen Patients in Colombia, using Deep Learning Architectures as YoloV7 and U-Net, utilizing Python and TensorFlow. Applied the framework to track chronic wounds evolution and assist medical treatment.
● Processed and Labeled 300+ Chronic Wound Images for detection and binary segmentation tasks, using Label-Studio. Processed Images were used in the training of a Deep Learning Framework.
● Developed a Deep Learning Unrolled Algorithm for Phase Retrieval, using TensorFlow, obtaining +5 dB improvement in Peak-to-Signal Noise Ratio for single snapshot Phase Retrieval.
● Designed a custom activation function for Binary Neural Networks, using MATLAB, TensorFlow and Pytorch, obtaining 90% accuracy in the classification tasks on datasets as CIFAR-10 and STL-10.
Software Engineer Degree from Universidad Industrial de Santander. Obtained the Cum Laude distinction.
Some moments I want to share
International Workshop on Adaptive, Compressive and Computational Imaging (Wacci). It was an international event in the university, organized by HDSP group where researchers from Chile, Uruguay, United States, Japan, Spain and Peru participated, the event lasted 3 days and one of that days I presented one of the projects I was working on to the community.
A day when HDSP students presented some of the work we did to the community of the university, there were students of many careers.