Project :Intelligent Environmental Risk Assessment in Oil Well Safety: Leveraging Advanced NLP and AI Techniques.Spearheaded critical data extraction from PDF documents related to environmental safety assessments of oil wells, employing advanced Natural Language Processing (NLP) techniques.Leveraged Hugging Face's cutting-edge NLP models and Transformers (BERT) for sentiment analysis on extracted data, contributing pivotal insights for environmental risk evaluation.Demonstrated… Show more Project :Intelligent Environmental Risk Assessment in Oil Well Safety: Leveraging Advanced NLP and AI Techniques.Spearheaded critical data extraction from PDF documents related to environmental safety assessments of oil wells, employing advanced Natural Language Processing (NLP) techniques.Leveraged Hugging Face's cutting-edge NLP models and Transformers (BERT) for sentiment analysis on extracted data, contributing pivotal insights for environmental risk evaluation.Demonstrated adaptability to industry-specific tasks and AI technologies, showcasing proficiency in Language Model (LLM) implementation, Generative Adversarial Networks (GANs), and expert data extraction within the challenging context of oil well safety assessments.Applied fine-tuning methodologies using LORa and QLora to customize models according to project requirements, optimizing performance and ensuring robust results.Orchestrated Falcon 7B dataset structuring and library setup, incorporating HuggingFace Transformers, Datasets, and WandB for streamlined training progress monitoring.Selected and configured Falcon 7B LLM model, defined PEFT parameters for LoRA, and implemented quantization strategies, balancing memory efficiency with acceptable error rates.Defined training arguments, including batch size, optimizer, learning rate scheduler, and checkpoints, for the fine-tuning process.Executed fine-tuning using the HuggingFace Trainer with PEFT configuration, monitored training progress with WandB, and maintained a vigilant approach to prevent overfitting through continuous validation of both training and validation loss Show less

Client : Henry SchienProject : Supplier Lead Time PredictionLed a team of 5+ data scientists and data engineers to develop a regression algorithm for predicting the number of days a supplier would take to send a product to Henry Schein's warehouse.Utilized Google Cloud Platform tools such as BigQuery, SQL, and Python libraries, including Keras and TensorFlow, to implement a smart algorithm, significantly enhancing warehouse management by accurately forecasting supplier lead… Show more Client : Henry SchienProject : Supplier Lead Time PredictionLed a team of 5+ data scientists and data engineers to develop a regression algorithm for predicting the number of days a supplier would take to send a product to Henry Schein's warehouse.Utilized Google Cloud Platform tools such as BigQuery, SQL, and Python libraries, including Keras and TensorFlow, to implement a smart algorithm, significantly enhancing warehouse management by accurately forecasting supplier lead times.Conducted Pareto analysis on a vast dataset comprising 400K customers, suppliers, and 3000 products, identifying high- profit and frequently ordered items for focused optimization.Investigated the factors influencing certain variable(s) in different periods, contributing to a more nuanced understanding of the supplier lead time prediction model.Executed regression analysis to refine and optimize the predictive model, emphasizing precision and reliability in predicting supplier lead times for 400K customers and suppliers dealing with 3000 products.Analyzed the historical dataset and its patterns to comprehend the underlying dynamics influencing supplier lead times. Matched the current situation with patterns derived from the previous stage, ensuring alignment between historical trends and real-time data for enhanced accuracy.Contributed to a deeper understanding of customers, streamlined warehouse operations, and refined business strategies, ultimately enhancing Henry Schein's overall operational efficiency and customer satisfaction. Show less

Client : HONG LEONG BANK MALAYSIAProject : Customer Segmentation and Revenue Growth Prediction for Hong Leong Bank MalaysiaLed and coordinated two teams of data engineers and ML engineers in the development of a sophisticated customer segmentation algorithm using K-Means clustering within the Google Cloud Platform (GCP) infrastructure. Implemented advanced techniques within the K-Means clustering process to enhance accuracy and granularity in customer segmentation.… Show more Client : HONG LEONG BANK MALAYSIAProject : Customer Segmentation and Revenue Growth Prediction for Hong Leong Bank MalaysiaLed and coordinated two teams of data engineers and ML engineers in the development of a sophisticated customer segmentation algorithm using K-Means clustering within the Google Cloud Platform (GCP) infrastructure. Implemented advanced techniques within the K-Means clustering process to enhance accuracy and granularity in customer segmentation. Managed large-scale customer datasets, overseeing projects focused on Motor Insurance cross-selling, Customer Segmentation based on spending patterns, and Revenue Growth prediction through the utilization of various bank products.Conducted data preprocessing using Google BigQuery and Vertex AI, implementing Python libraries such as Pandas and Numpy for efficient data handling and manipulation.Developed and fine-tuned machine learning algorithms, incorporating regression models for Revenue Growth prediction. Applied statistical methods and feature engineering to optimize the accuracy and reliability of the models. Spearheaded data migration initiatives to streamline processes and enhance overall data efficiency.Designed and implemented an interactive Google Analytics dashboard, providing stakeholders with a visually intuitive platform for data exploration and insights.Employed advanced ML techniques to enhance Revenue Growth prediction models, leveraging insights derived from spending patterns and customer segmentation.Collaborated with business stakeholders to understand objectives and ensure ML algorithms aligned with strategic goals, facilitating effective cross-functional teamwork.Contributed to Hong Leong Bank Malaysia's data-driven decision-making process, enabling personalized customer engagement, targeted marketing efforts, and sustainable revenue growth. Show less

Developed ML-Spark scripts processing thousands of images for prediction algorithm.Created synthetic data using Keras data augmentation for better training of the model.Developed a computer vision algorithm with Convolution Neural Networks (CNN's) TensorFlow and Keras for cattle recognition using the SIFT (Scale-Invariant Feature Transform) technique.Conducted experimental evaluations, demonstrating the superior performance of the proposed algorithm.Developed a generative AI… Show more Developed ML-Spark scripts processing thousands of images for prediction algorithm.Created synthetic data using Keras data augmentation for better training of the model.Developed a computer vision algorithm with Convolution Neural Networks (CNN's) TensorFlow and Keras for cattle recognition using the SIFT (Scale-Invariant Feature Transform) technique.Conducted experimental evaluations, demonstrating the superior performance of the proposed algorithm.Developed a generative AI model using GAN's which can generate high-quality natural images that develop gradually to generate more and more realistic looking data by coupling with an adversarial network.This framework not only has the possibility of generating very high-quality synthetic data but also it can be used to enhance pixels in photos,conversion of images from one domain to another.Achieved a high identification accuracy of 93.3% within a reasonable processing time.Outperformed traditional identification approaches, which achieved an identification accuracy of 84%. Show less

Client: RGE GroupDeveloped a CNN project using Python to identify optimal tree crowns for paper-making in a paper mill. Applied convolutional neural network algorithms to determine suitable cutting points, enhancing efficiency in paper production. Implemented a precise algorithm, enhancing the paper-making process by automating the identification of ideal tree crown cutting points using convolutional neural networks in Python.Employed a Python-based machine learning recommender model… Show more Client: RGE GroupDeveloped a CNN project using Python to identify optimal tree crowns for paper-making in a paper mill. Applied convolutional neural network algorithms to determine suitable cutting points, enhancing efficiency in paper production. Implemented a precise algorithm, enhancing the paper-making process by automating the identification of ideal tree crown cutting points using convolutional neural networks in Python.Employed a Python-based machine learning recommender model using Artificial Neural Networks and Random Forest algorithm to predict KAPPA values for RGE Group Indonesia, incorporating TensorFlow and Keras frameworks to address the complexity of the task.Collected and cleansed data using SQL(ETL), conducted a decade-spanning data analysis to anticipate optimal KAPPA values based on diverse parameters.Enhanced data visualization through a Tableau dashboard, while also crafting a deep learning model for precise KAPPA number predictions. Show less

In the project "Caterpillar Engine Image Detection Using CNN" (Convolutional Neural Networks), my tasks included Exploratory data analysis for pipeline establishment, Choosing a network architecture and experimenting with design,Exploring pre-processing techniques to enhance model performance and Utilizing mind maps for process optimization and continuous improvement.Created synthetic data(Engine Images) using Keras data augmentation for better training of the model.The main achievement… Show more In the project "Caterpillar Engine Image Detection Using CNN" (Convolutional Neural Networks), my tasks included Exploratory data analysis for pipeline establishment, Choosing a network architecture and experimenting with design,Exploring pre-processing techniques to enhance model performance and Utilizing mind maps for process optimization and continuous improvement.Created synthetic data(Engine Images) using Keras data augmentation for better training of the model.The main achievement of the project was developing an algorithm to predict engine model numbers from provided images. For the same client, Caterpillar, I developed an additional machine learning algorithm for predictive maintenance of boat engines. This involved: Collecting data from various Engine Control Units (ECUs) installed on the engine, Employing Artificial Neural Networks to address this intricate challenge, Implemented predictive maintenance analytics using machine learning models built with Python. Show less

Client: Siloam Hospitals(Indonesia)As a data analyst for the project "Demand Planning and Inventory Management" at Siloam Hospitals, my responsibilities included:Extracting raw data and developing a Data Discrepancy report across different data sources,Migrating data from MySQL to Microsoft Excel Sheets and further processing it in Python for analysis.Used NLP in demand forecasting by analyzing text data for future demand.Utilizing the data discrepancy report to perform Pareto… Show more Client: Siloam Hospitals(Indonesia)As a data analyst for the project "Demand Planning and Inventory Management" at Siloam Hospitals, my responsibilities included:Extracting raw data and developing a Data Discrepancy report across different data sources,Migrating data from MySQL to Microsoft Excel Sheets and further processing it in Python for analysis.Used NLP in demand forecasting by analyzing text data for future demand.Utilizing the data discrepancy report to perform Pareto Analysis,classifying SKUs into top 70%, mid 20%, and low 10% categories based on their value and profitability, using SQL, Python, Scikit-Learn, Pandas, Matplotlib, Seaborn and Power BI. Show less