YOGYAKARTA – Turning concepts from the classroom into real-world implementation, the AI Talent Factory (AITF) program—a collaborative initiative between the Ministry of Communication and Digital (Komdigi) and Universitas Gadjah Mada (UGM)—successfully held its Workshop III on June 2-3, 2026. Focusing on the refinement of the Minimum Viable Product (MVP), this crucial phase served as a proving ground for technical readiness and system integration by four teams of artificial intelligence student talents.
The highlight of the workshop was the demonstration session, where Team 4, representing their peers, successfully presented the interactive and ready-to-use KPM x AITF Dashboard to the Director of Public Information and the Director of Public Communication and Media (KPM) of Komdigi.
Broadly speaking, this smart dashboard operates automatically, starting from crawling text, image, and audio data from social media and online platforms. Through AI technology, the raw data is then enriched using image captioning models, audio transcription, and sentiment analysis to generate a final product in the form of issue narratives and Daily Briefs.
The success of this system integration is inseparable from the remarkable technical leaps achieved by each team:
-
Team 1 (Infrastructure and Data Optimization): Successfully integrated the keyword manager, improved the crawler workflow, and strengthened the model's infrastructure through Supervised Fine-Tuning (SFT) for images and text using the Qwen model, specifically optimizing image captioning.
-
Team 2 (Model Processing Efficiency): Successfully achieved the Continual Pre-Training (CPT) target of up to 270 million dataset tokens. This team also drastically reduced the model training time from 18 hours to just 8 hours by leveraging Unsloth efficiency, recording a training time efficiency of up to 55%.
-
Team 3 (API Integration and Regulatory Evaluation): Played a vital upstream role by ensuring the stability of the API system for data-based Daily Brief generation (in JSON format). The team also equipped the system with Retrieval-Augmented Generation (RAG) technology to evaluate the relevance of related regulatory documents and legal articles.
-
Team 4 (Visualization and User Interface): Successfully became the culmination of innovation by integrating all technical components from the other teams into an interactive monitoring Dashboard ready for stakeholder use.
Responding to the impressive demonstration, the Director of Public Information and the Director of KPM provided strategic feedback. Future development is no longer solely focused on technology but is now directed towards more detailed issue filtering. Furthermore, the system is challenged to possess highly capable predictive abilities to detect and forecast potentially viral issues in the future (what next will happen).
With continuous evaluation and refinement underway, these proud national AI talents are increasingly prepared to face the next challenges. The entire technical process will culminate in final testing at Workshop IV and the upcoming Demo Day event scheduled for late June.