Rio Julian Logo Image
Rio Julian
hamburger menu hamburger menu close

Hi There, I'm Rio👋

Welcome to my portfolio page! Here, you'll find a collection of the projects that I've worked on as a Data Scientist. You can also find my resume by hitting the button below. Enjoy your exploration!

Resume

About Me

Get to know me!

Hey! I'm Rio and I'm a Data Scientist with 3 years experience who love to work with data and extracting the valuable insights behind it. I'm a Computer Engineering graduate from Universitas Diponegoro with cum laude GPA (3.67/4.00). As a data scientist, I have strong knowledge in machine learning, generative AI, data analysis, and data visualization.

I have experience in productionizing machine learning models on cloud platform environment (Google Cloud Platform and Amazon Web Services) using MLOps lifecycle which contains various steps from data collection, model deployment and monitoring the deployed machine learning model. Other than that, I also have experience in building generative AI application using state-of-the-art large language models and their supporting services or frameworks.

Currently I'm working at SG-EDTS, which is Salim Group subsidiaries that provides digital technology and analytics consultancy, particularly for Salim Group companies. You can find their website here.

Personally, I love to learn new things and share good things with enthusiasm and a high ambition to achieve goals. Feel free to contact me here.

Contact

Technical Skills

Python
SQL
Tableau
Google Data Studio
Pandas
Numpy
Scikit Learn
Tensorflow
Keras
Pytorch
Docker
Kubeflow
API
Google Cloud Platform
Azure Machine Learning Platform
AWS Sagemaker
ETL
LangChain
Prompt Engineering
Chroma DB

Projects As a data scientist, I've worked on various projects, whether it is descriptive or predictive analytics. Because currently I work in SG-EDTS, which is a Salim Group Data and IT consultant company, I have experienced to handle project from various Salim Group company. Below are the projects I've worked on so far:

Product Recommendation

Klik Indomaret - Product Recommendation System

Klik Indomaret is an E-Commerce which sells more than 5.000 unique products. I've helped them to create personalize product recommendation based on their customer purchase behaviour. In this project, I was responsible to handle various task, that is:
• Designed and developed machine learning pipeline with Vertex Pipeline (Kubeflow Based)
• Build inhouse product recommendation model (from previous AWS personalize services) which could reduce the machine learning cloud services costs up to 90%
• Experimented with different algorithm such as: SVD, Autoencoder-based model, Word2Vec, Transformer
• Processed the data, feature engineering and evaluated the product recommendation model with various metrics (Precision, Recall, NDCG)
• Tested the API with JMeter (concurrent request simulation)

Technology Used:


Google Vertex AI
Google Cloud BigQuery
Google Cloud Storage
Google Cloud Run
Google Cloud Build
Google Cloud Endpoints
Google Cloud Container/Artifacts Registry
Kubeflow Pipeline
Docker
Tensorflow
Keras
Pytorch
Pandas
FastAPI
KTP Detection

Gurih - KTP Image Classification & Localization

Gurih is an application that can be used by minyak curah seller to make transactions and controlling goods easier. With limited supplies, the government requires that buyer should use their ID Card (KTP Images) to bought the Minyak Curah, but sometimes the buyer or seller ignore that regulation. To overcome that problem, me and my team create a machine learning model that could classify ID Card (Real, Photocopied & None) and could detect them. You can find the example of ID Card classification and detection on here. In this project, I was responsible to handle some task, such as:
• Developed ID Card detection model using pre-trained FasterRCNN Inception ResNet V2 model
• Handled prediction pipeline, developing and integrating various API with Google Cloud Run
• Evaluated the performance of the model
• Explored the ID Card feature analysis to alleviate performance of the classification model

Technology Used:


Google Vertex AI
Google Cloud Storage
Google Cloud Build
Google Cloud Run
Docker
Tensorflow
Keras
Pandas
OpenCV
FastAPI
Generative AI Application

Various Clients - Chatbot Powered by Generative AI

We build chatbot application powered by state-of-the-art large language models such as ChatGPT 4 and Gemini Pro for our various clients such as Klik Indomaret, Toyota Astra Motors and Kalbe Consumer Health. The use case of our developed chatbot can vary depending on the intended purpose such as: customer support agent, data analysis agent (table or chart image-based) and information retrieval agent with human-like conversation. In this project, I was responsible to handle some task, such as:
• Developed chatbot agent, created conversational flow and their intents with Dialogflow CX
• Integrated the chatbot agent with large language models (Google Gen AI and Azure Open AI ChatGPT) via cloud function webhook
• Performed the optimal prompt and model parameter for various use case of our chatbot (prompt engineering)
• Build the multi-modal retrieval augmented generation (RAG) which consists of text, table and image data using LangChain framework, Chroma DB and Vertex Search AI
• Developed the UI chatbot web-apps application using JavaScript and CSS bootstrap framework and deployed it with Google App Engine

Technology Used:


Google Dialogflow CX
Google Generative AI: Gemini Pro, Text-Bison, Text-Unicorn
Azure Open AI ChatGPT 3.5 & 4
Google Vertex Search AI
Google Cloud Function
Google App Engine
LangChain
Chroma DB (Vector Database)
Cluster Analysis

Indogrosir - Customer Clustering Analysis

Indogrosir is a retail company in the modern wholesale retail sector that provides basic daily needs. With more than 200K users across Indonesia, me and my team initiate to conduct clustering analysis to find their customer hidden characteristics. In this project, I was responsible to:
• Created user and item catalog, and collect various features related to user and item
• Experimented to create cluster with KMeans Algorithm
• Evaluated the clustering model results
• Provided analysis deck and delivered the results to the stakeholders

Technology Used:


Google Vertex AI
Google Cloud BigQuery
Google Cloud Scheduler
Scikit Learn
Pandas
Numpy
Seaborn
Loss Item Prediction

Indomaret - Loss Item Prediction

Indomaret, a widely-known retail chain with over 21,900 stores across 32 provinces in Indonesia, deals with typical issues related to lost items. These losses often occur due to theft (from both inside and outside the company), mistakes in administration, and damaged goods. Our goal is to assist them by providing estimates for the total number of lost items in Indomaret stores. This facilitates their team in prioritizing which stores require inspection first.
In this project, my responsibilities included:
• Provided descriptive analytics utilizing Google BigQuery and Tableau
• Experimented with the development of forecasting machine learning models using LSTM and XGBoost regressors with varying window sizes
• Evaluated the performance of developed model
• Compiled the entire analysis deck and delivered the results to the stakeholders

Technology Used:


Google Vertex AI
Google Cloud BigQuery
Tableau
Scikit Learn
Pandas
XGBoost
Text Classification

EDTS (Research Project) - News Headlines Classification

News Headlines Classification Apps is an apps that could classify the Indonesian news headlines. This project was a part of my data management trainee main task in EDTS. I was with my data engineer friend when I built this apps, he handled the data collection via scraping and deployment things and I handled some tasks such as:
• Developed text classification model with LSTM and Word2Vec (as a feature extractor) algorithm
• Evaluated the performance of the model with confusion matrix
• Handled the front-end part, implemented visual and interactive elements on the web apps with HTML and CSS Bootstrap

Technology Used:


Google Colab
NLTK
Word2Vec
Tensorflow
Keras
Pandas
Numpy
FastAPI
HTML
CSS
Text Classification

Indofood CBP - Business Performance Analysis

PT Indofood CBP Sukses Makmur Tbk is a leading Indonesian consumer branded products company, specializing in a diverse range of food and beverage categories. This is my descriptive analytics project which should provided the monthly business performance for Indofood CBP principles/brand. Whereas my scope in this project are:
• Prepared the datamart which collected from the dataset in Google BigQuery
• Analyzed data with Microsoft Excel and Tableau, and visualize it with Thinkcell
• Created query and pivot table frameworks which could made analysis more easier (semi-automate)
• Helped team to provide the business performance deck for 9 principle (Noodle, Nutrition, Beverage, etc.)

Technology Used:


Google BigQuery
Tableau
Thinkcell
Microsoft Excel
Microsoft Powerpoint

Contact Open for new opportunities in Data or Machine Learning Field. For any further Information, please contact me below: