Selected projects

Turning technical possibility into working systems.

A selection of AI, recommendation, forecasting, and analytics work across media, consulting, industrial, and enterprise settings.

RecommendationVision+ · Media & Entertainment

Multi-Model Recommendation Engine

A portfolio of recommendation models designed to make content discovery more timely, personal, and relevant.

View project story

Challenge

Turn large-scale user behavior, customer profiles, and content metadata into fast, relevant recommendations for a streaming platform.

Approach

Designed Trending, Popular, Seasonal, and Because You Watched models, plus content similarity using Indonesian text preprocessing and cosine similarity.

Outcome

Delivered recommendations from millions of behavioral and profile records in under two minutes, supported by A/B testing and model documentation.

AI & AgentsAll Data International · Generative AI

Enterprise Knowledge AI Agent

An agent architecture that retrieves answers from documents, tabular files, and SQL sources through one conversational interface.

View project story

Challenge

Make fragmented enterprise information easier to access across unstructured files and structured data sources.

Approach

Designed the agent architecture, RAG flow, vector retrieval, prompt routing, and integration approach for PDF, DOCX, TXT, PPTX, CSV, Excel, and SQL data.

Outcome

Created a reusable product foundation for grounded enterprise question answering and faster access to internal knowledge.

ForecastingEnterprise Client · Industrial Analytics

Commodity & Operational Forecasting

Forecasting models for commodity, marine, and weather variables designed around operational decision needs.

View project story

Challenge

Provide decision-ready forecasts across coal, LNG, natural gas, crude palm oil, marine, and weather data.

Approach

Created and fine-tuned Prophet models, designed the end-to-end data flow, and aligned testing and model outputs with client requirements.

Outcome

Delivered an extensible forecasting capability covering multiple business-critical variables and operational contexts.

RecommendationOkezone & BuddyKu · Digital Media

News Personalization at Scale

Keyword and behavioral recommendation systems developed across production media environments.

View project story

Challenge

Improve content relevance while keeping personalization algorithms efficient across active news products.

Approach

Led AI engineering squads, designed recommendation and pipeline architecture, and used KeyBERT, YAKE, tags, keywords, and user behavior signals.

Outcome

Improved one user-history algorithm's processing efficiency by up to 50%, with A/B tests across UAT, staging, and production.

AI & AgentsCoal Industry · Decision Intelligence

News Sentiment Intelligence Agent

An AI agent concept combining article analysis, sentiment intelligence, and business-ready reporting.

View project story

Challenge

Transform high-volume news coverage into a more structured view of sentiment and potential business relevance.

Approach

Designed the AI agent system, architecture, development scope, and cost model for sentiment analysis over news articles.

Outcome

Established a practical technical and commercial blueprint for an AI-enabled news monitoring capability.

AnalyticsAll Data International · Internal Product

Data Management Intelligence Tools

A suite of algorithms for exploration, cleansing, geocoding, reverse geocoding, and area similarity.

View project story

Challenge

Standardize reusable data preparation and location intelligence capabilities for downstream product teams.

Approach

Led architecture, code reviews, GitLab delivery, documentation, task planning, and hands-on Python development.

Outcome

Delivered a structured algorithm foundation with documented interfaces for backend integration and continued product development.

AnalyticsAll Data International · Internal Analytics

Enterprise 360 Performance Dashboard

A cross-department monitoring and analytics flow built from ClickUp and Google Sheets data.

View project story

Challenge

Create a consistent source of decision-ready performance data from multiple operational platforms and formats.

Approach

Led the data ingestion, cleansing, architecture, visualization, quality standards, and technical coordination across Data Science, Analytics, and PMO.

Outcome

Produced a standardized data flow and dashboard foundation for clearer goal-based monitoring across departments.

AnalyticsEnergy Analytics

Electricity Usage Anomaly Detection

An analytical model designed to surface unusual consumption behavior across multiple customers.

View project story

Challenge

Identify meaningful deviations in electricity consumption patterns across diverse customer profiles.

Approach

Structured the anomaly detection use case around customer-level usage patterns and interpretable analytical outputs.

Outcome

Created a reusable analytical basis for investigating unusual energy consumption and prioritizing follow-up analysis.

A note on confidentiality

Project stories focus on scope, methods, and verified outcomes from my CV. Sensitive client data, proprietary implementation details, and confidential commercial information are intentionally excluded.