My Projects

FinFlow AI — Automated Personal Cashflow System

Tools: Artificial Intelligence, Automation, n8n, Spreadsheet, Python, Looker

Description: This project builds an automated personal cashflow management system using n8n, enabling seamless expense and income tracking from multiple input sources. The workflow automates data capture from receipts using AI, parses financial notifications from Gmail, supports manual input via Telegram Bot, and stores results in Google Sheets for visualization in Looker Studio. It also includes an AI assistant on Telegram to analyze financial patterns and provide personalized insights.

Steps:

• Configure n8n workflow to capture cashflow data from three entry points: 1. Receipt scanning & OCR with AI; 2. Gmail financial transaction notifications; 3. Manual inputs via Telegram Bot.
• Process and clean extracted data for structured recording.
• Store all transactions into Google Sheets as a centralized data source.
• Connect Looker Studio for dashboards and financial insights.
• Integrate an AI assistant in Telegram for real-time financial Q&A and spending advice.

Results: Successfully delivered an automated personal finance system that reduces manual bookkeeping, centralizes cashflow data from multiple sources, and provides real-time visibility into financial behavior. Users can scan receipts, log spending through chat, and ask AI for personalized budgeting suggestions based on historical patterns.

Streamlit Dashboard on Snowflake

Tools: Streamlit, Snowflake, Python, Data Visualization

Description: This project builds a custom dashboard using Streamlit on top of Snowflake, providing an interactive platform for business users and analysts to visualize data directly from Snowflake.

Steps:

• Connect Streamlit application with Snowflake warehouse.
• Design and develop dashboards with interactive filters, charts, and metrics.
• Optimize queries for real-time data visualization.
• Deploy dashboard for team and stakeholder usage.

Results: Delivered an interactive, user-friendly dashboard that leverages Snowflake’s data warehouse performance with the flexibility of Streamlit for visualization.

BigQuery Approval Management

Tools: BigQuery, GitLab, Data Governance

Description: This project establishes a framework to connect BigQuery with GitLab, enabling version control, review processes, and stronger data governance for SQL scripts and data models. All changes are tracked to ensure security, accountability, and compliance. The system also includes an audit mechanism to monitor all query activities and restrict high-risk commands such as DROP, DELETE, and UPDATE without proper approval.

Steps:

• Integrate GitLab CI/CD with BigQuery environment.
• Enable peer-review process for SQL queries and schema updates.
• Automate deployments to production with approval gates.
• Implement an audit system to log all query activities in BigQuery.
• Ensure version history and rollback capabilities for queries.

Results: Improved data governance, safer production deployments, full query activity auditing, and minimized risk of destructive changes through strict approval control and transparent review processes across the data team.

DBGemmy - Generator Database Dummy

Tools: Artificial Intelligence (AI), Python, Web Development

Description: Developed an AI-powered web application called DBGemmy that generates realistic dummy databases using Gemini API and various Python concepts. This tool automates the creation of sample datasets in table format, enabling developers and analysts to quickly simulate data for testing, prototyping, and demonstration purposes.

Steps:

• Needs Analysis: Identified the need for customizable dummy data to support development, testing, and learning use cases.
• Technology Integration: Combined various Python concepts with Gemini’s AI API to generate diverse and context-aware data fields.
• Web Development: Built a user-friendly web interface for table generation using Flask/Streamlit.
• Dynamic Input Handling: Implemented logic for users to define the context, column names, data types, and number of rows.
• Data Generation: Enabled real-time generation of structured dummy data in tabular format.
• Output & Export: Added functionality to display data on the web and optionally export it to CSV, Excel, or other formats (soon).

Results: Successfully launched DBGemmy, a practical AI-driven dummy data generator that simplifies data creation workflows for developers. The tool has enabled faster prototyping and more realistic testing environments through automated, customizable dummy datasets.

Promo Affinity Project at LinkAja

Tools: Automation, Excel, BigQuery, Looker, Confluence

Description: Implemented the Promo Affinity Project at LinkAja to enhance targeted marketing by analyzing user behavior, preferences, and products. The project aimed to optimize promotional strategies, improve user engagement, and increase conversion rates through data-driven insights.

Steps:

• Assessment Data: Conducted assessments on the related services to ensure the required data is available and meets business needs.
• Data Collection & Analysis: Aggregated and analyzed transaction data in BigQuery to identify user affinity patterns.
• Segmentation: Created dynamic user segments based on purchase behavior and engagement levels.
• Visualization: Created real-time dashboards in Looker to monitor promo performance and user engagement.
• Documentation: Maintained detailed records in Confluence for cross-team accessibility.

Results: Enabled targeted promotions with improved accuracy, boosted user engagement, and enhanced promotional effectiveness through precise data segmentation and real-time performance monitoring.

Automation of Flagging System at LinkAja

Tools: Automation, Excel, BigQuery, GitLab, Looker.

Description: Developed and implemented an automated flagging system in the LinkAja app to differentiate between “whitelisted” and “blacklisted” users who are eligible for further actions, such as promotion flagging, target disbursement, new feature target, and more, thereby eliminating manual processes and enhancing efficiency.

Steps:

• Data Assessment & Collaboration: Worked with relevant teams to analyze data needs.
• Procedure Development: Searched for source tables in BigQuery and built temporary tables for non-BigQuery data, then developed procedures in BigQuery.
• Automation: Integrated the process with Airflow for full automation.
• Monitoring & Visualization: Set up real-time monitoring in Looker.
• Validation & Testing: Ensured system accuracy and reliability.
• Documentation: Maintained detailed records in Confluence for cross-team accessibility.

Results: Achieved nearly 100% operational efficiency, minimizing manual tasks and providing accurate, real-time reporting.

Automation of Fraud Checking System at Bukalapak

Tools: Automation, Fraud Analysis, Spreadsheet, BigQuery, Python, Looker Studio.

Description: Developed and implemented an automated fraud checking system at Bukalapak to enhance efficiency, accuracy, and reduce potential losses from fraudulent transactions.

Steps:

• Needs Analysis: Collaborated with product and relevant teams to understand existing processes.
• System Design: Integrated Spreadsheet, BigQuery, Python, GitLab, and Looker Studio, utilizing Pub/Sub, Docker, and Airflow.
• Implementation: Automated data processing in BigQuery, applied fraud detection logic in Python, and visualized results in Looker Studio for real-time monitoring.
• Optimization: Improved system performance for handling large data volumes efficiently.
• Validation & Testing: Ensured system accuracy and reliability.
• Documentation: Maintained detailed records in Confluence for cross-team accessibility.

Results: Delivered a robust system that streamlined fraud detection, minimized manual tasks, provided real-time, accurate reporting, and enhancing operational efficiency up to 90%.

Automation of GCP Hardening Checking System at Bukalapak

Tools: Automation, Data Security, Spreadsheet, BigQuery, Python, GitLab.

Description: Developed and implemented an automated security checking (hardening) system for Bukalapak’s Google Cloud Platform (GCP) to enhance security and ensure compliance with internal policies and external regulations (e.g., UU PDP, GDPR).

Steps:

• Needs Analysis: Collaborated with product, infrastructure, and compliance teams to identify security gaps and regulatory needs in GCP.
• System Design: Designed a modular system to assess GCP configurations against security benchmarks and policy standards.
• Implementation: Developed Python-based automation integrated with GitLab CI/CD to continuously scan and evaluate security settings.
• Optimization: Improved performance through API batching, metadata caching in BigQuery, and automated alerting/reporting workflows.
• Validation & Testing: Conducted functional testing across environments and validated results with the security team for accuracy.
• Documentation: Documented implementation, rule logic, and response procedures to ensure maintainability and compliance.

Results: Delivered a proactive system for continuous security monitoring, early risk detection, and improved compliance, significantly boosting security practices in Bukalapak’s GCP environment.

Tech Annual Risk Assessment Project at Bukalapak

Tools: Data Security, Risk Assessment, Archer, Spreadsheet.

Description: Conducted annual risk assessments within the technology domain at Bukalapak. Assisted multiple business units, including Buka Investasi Bersama, BukaSend & Logistics, Lapak Gaming, Financing Marketplace, and the Data Engineering Team — in carrying out their annual risk assessments.

Steps:

• Led a structured risk assessment process at Bukalapak.
• Involving coordination with business units
• Identification and analysis of technology-related risks using the RSCA format from Archer.
• Development of mitigation strategies, comprehensive documentation via the Archer Dashboard.
• Continuous follow-up to monitor risk mitigation progress.

Results: The project provided clear insights into technological risks across Bukalapak's business units and led to effective mitigation strategies, resulting in stronger system resilience, better regulatory compliance, and improved risk management.

Web Scraping Project for Indonesian Postal Code Data

Tools: Web Scraping, ETL, Python, BigQuery, Looker

Description: This project aims to automatically collect postal code data from the Kodepos Indonesia website. This data is needed to enrich geographical information such as provinces, cities, districts, sub-districts, and postal codes, which are used for various business data analysis and processing needs.

The data collection process is carried out using web scraping or API retrieval techniques, and the results are processed into table format in BigQuery. The project focuses on efficiency, accuracy, and ensuring the collected data meets the specific needs of the users.

Definition of Ready (DoR):

• Create Syntax to Scrape Website
• Scrape Data from Website

Definition of Done (DoD):

• Sync and update to datamart geo_map

AGROSTOCK: Fruit Freshness Detection App Using a CNN Model

Tools: Machine Learning, Python.

Description: This project developed a CNN-based image processing model to assess fruit freshness based on visual appearance and suitability for consumption.

Steps:

• Data Collection: Collected a labeled dataset of fruit images categorized by freshness levels.
• Image Preprocessing: Applied resizing, normalization, and data augmentation to improve dataset quality and diversity.
• Model Construction: Built a Convolutional Neural Network (CNN) model architecture suitable for image classification tasks.
• Model Training: Trained the CNN model using the processed dataset to recognize patterns related to fruit freshness.
• Model Validation: Tested the model on unseen data to evaluate its accuracy in classifying fruit freshness levels.
• Optimization and Fine-Tuning: Improved the model’s performance through hyperparameter tuning and optimization techniques.
• Implementation: Deployed the trained model into an application to detect fruit freshness easily and efficiently.

Results: This project developed a machine learning model to assist in detecting fruit freshness, with its accuracy and performance thoroughly evaluated for reliability.

COVID-19 Detection Using X-Ray Images and Machine Learning Models

Tools: Machine Learning, Python.

Description: This project develops a CNN-based machine learning model to detect potential COVID-19 cases by analyzing patterns in X-ray images.

Steps:

• Data Collection: Gathered a dataset of X-ray images including COVID-19 positive, negative, and other lung conditions like pneumonia or normal cases.
• Image Preprocessing: Normalized pixel intensity and resized images to prepare the data for modeling.
• Model Construction: Built a CNN architecture designed for detecting COVID-19 from X-ray images.
• Model Training: Trained the CNN model to recognize patterns linked to COVID-19 in the preprocessed images.
• Model Validation: Tested the model on unseen data to evaluate its ability to distinguish between COVID-19 positive and negative cases.
• Optimization and Fine-Tuning: Improved the model's accuracy and generalization through hyperparameter tuning and architecture adjustments.

Results: This project developed a machine learning model to assist in detecting potential COVID-19 cases from X-ray images, with its accuracy and performance thoroughly evaluated for reliability.

Sentiment Analysis of Public Responses to the Semarang City Health Department

Tools: Sentiment Analysis, Web Scraping, Machine Learning, Python, Streamlit.

Description: Conducted annual risk assessments within the technology domain at Bukalapak. Assisted multiple business units, including Buka Investasi Bersama, BukaSend & Logistics, Lapak Gaming, Financing Marketplace, and the Data Engineering Team — in carrying out their annual risk assessments.

Steps:

• Data Collection: Scraped comment data from Dinkes Semarang’s Instagram posts, covering positive, negative, and neutral sentiments.
• Data Preprocessing: Processed the text data through tokenization, stopword removal, and stemming/lemmatization to prepare for modeling.
• Model Building: Developed multiple sentiment classification models, including TF-IDF, Word2Vec, CNN, Random Forest, SVM, and LSTM.
• Model Training: Trained models on the preprocessed dataset to detect sentiment patterns effectively.
• Model Evaluation: Assessed model performance using accuracy, precision, recall, and F1-score.
• Optimization: Fine-tuned models to enhance performance and generalization.
• Deployment: Integrated models into Streamlit and Gradio for real-time sentiment analysis of new Instagram comments.

Results: This project produced machine learning models to analyze public sentiment on Instagram related to the Semarang City Health Department, providing insights into perceptions of local health services.

Other Projects: Building Scalable Analytics and Governance Frameworks

• ⚙️ Data Pipeline & Automation
  Automated ingestion, transformation, and merge-upsert processes in GCP and Snowflake. Reduced manual intervention and improved data freshness through scheduled pipelines.
  Key Outcomes: Increased operational efficiency and consistency in daily data processing.
  Example tasks: Automated onboarding/offboarding workflows for new data pipelines; Scheduled ingestion processes in Dataform and Cloud Composer; Developed merge-upsert scripts for Digipos and SPBU datasets.
• 📊 Data Visualization & Looker/Analytics
  Designed and maintained 70+ dashboards in Looker to support product, marketing, and operations analytics. Built reusable LookML models for unified metric tracking.
  Key Outcomes: Improved data-driven decision-making and cross-team visibility of KPIs.
  Example tasks: Created dashboards for transaction performance and user retention; Reviewed LookML metrics for consistency; Developed BAU dashboards for business stakeholders.
• 🧩 BigQuery Development & Data Modeling
  Improved data model design and optimized complex SQL queries to ensure scalability and maintainability. Supported data marts for product and merchant analytics.
  Key Outcomes: Enhanced query performance and improved reliability of analytical datasets.
  Example tasks: Optimized queries by partitioning and clustering; Created data warehouses and data marts for business purposes; Adjusted schema for several procedures (PRCs); Debugged and refined materialized views.
• 🔐 Data Governance & Access Control
  Implemented RBAC frameworks across Snowflake and GCP, ensuring compliance with GDPR, UU PDP, and internal data governance standards.
  Key Outcomes: Strengthened security posture and standardized access control processes.
  Example tasks: Created and managed all roles in LinkAja; Reviewed and updated access privileges; Standardized request and approval procedures.
• 🗂️ Documentation & Process Standardization
  Authored and maintained technical documentation for workflows, data repositories, and SOPs. Defined DoR (Definition of Ready) and DoD (Definition of Done) for Data Platform operations.
  Key Outcomes: Improved collaboration, onboarding efficiency, and audit readiness.
  Example tasks: Created documentation for various projects; Developed Snowflake onboarding guide; Standardized SOP templates for data change management.
• 🧹 Data Quality & Validation
  Designed a proof-of-concept framework for monitoring data quality across core datasets. Implemented validation and deduplication processes to ensure data integrity.
  Key Outcomes: Established baseline standards for continuous data quality improvement.
  Example tasks: Evaluated tools for Data Quality Framework; Designed validation scripts to detect duplicate data; Defined metrics for data completeness and accuracy.
• 🧠 Business-as-Usual (BAU) & Ad-hoc Analysis
  Delivered ad-hoc analytics and daily monitoring for product and business stakeholders, ensuring timely and reliable insights.
  Key Outcomes: Maintained business continuity through consistent and actionable data delivery.
  Example tasks: Extracted transaction data for JMTO and Digipos; Provided SQL analysis for fraud monitoring; Generated daily and weekly executive reports.