Data Scientist and AI Enthusiast
Innovative Data Scientist with expertise in implementing advanced AI solutions,
including hierarchical text classification, large language models, and optimization algorithms,
to drive efficiency and enhance decision-making.
Proven track record in demand forecasting, recommendation systems, and streamlining workflows, achieving significant operational improvements.
Skilled in leveraging data-driven strategies to deliver impactful business insights and scalable solutions.
Stack: Python, SQL, BigQuery, Metabase, PyTorch, Airflow, GCP, AWS
These are my personal project
In this project, we aim to improve the JakLapor service using state-of-the-art image classification in the hopes it will be able to reduce manual categorization for each report and as a category recommendation for the end-user reports they are taking.
How these models performance compare to each others with oversampling data using SMOTE (Synthetic Minority Oversampling Technique). how we handle data like this? we can use oversampling technique with minority class in order to make real-like data and finally the data has the same distribution.
How to teach drone to see what is below and segment the object with high resolution. Teaching drone to see is quite challenges due to bird’s eye view and most of pre-trained models are trained in normal images we see (point of view) in daily basis (ImageNet, PASCAL VOC, COCO).
This is semantic segmentation project to delineate salt bodies in seismic images with U-Net Architecture with Resnet18 and Resnet34. Salt can be a tricky process to image in seismic. To better interpret salt bodies in seismic, geologic information needs to be integrated in order to accurately image salt. Salt and sediment interactions can cause reservoir traps and hydrocarbon traps.
I classify 7 mineral classes and produce robust model that can be used to classify minerals, I've tried it using mineral images from Google.
Covid-19 is an unprecedented situation for society and it’s going to change how human lives for the next generation. It’s also making medical testing has to run quickly in massive test samples, which human can be tired and the machine does not, in a situation like this AI really help.
This Project about classify sentiment from scarapped twitter tweet using private data
Save your reading time while machine extracts important sentences and make summary from PDF file, using TF-IDF as weight for each word in sentences and return the highes score as summary and save your time.
This research uses predictive deconvolution and FK-filter to attenuate short period multiple from their move out, then continued by SRME method to predict multiple that cannot be attenuated from previous method, then followed by Radon transform to attenuate multiple that still left and cannot be attenuated by SRME method. The result of each method then compared to each other to see how well multiple attenuated.