Arya Nathan Bara

• Led a 30+ staff organization (500+ members) across six BINUS campus regions, overseeing 30+ initiatives through the year.
• Pitched and closed over IDR 100 million in corporate sponsorships (Paragon Corp, TikTok Shop, Tokopedia, and others).
• Grew Greater Jakarta recruitment registrants from 70 to 230 through rebranding, and raised the sign-up fee from IDR 75k to 175k (~IDR 40M in registration revenue).
• Launched an in-house talent agency connecting members to paid modelling and other job offers.

• Interviewed and selected ~200 Freshmen Leader / Partner candidates; coordinated administration and events with the Student Support Office.
• Directly mentored ~30 leaders, ensuring on-time completion of program deliverables and reports.

• Collaborated with cross-functional teams to develop and implement system enhancements.
• Delivered multiple SQL-based projects and provided user troubleshooting and support.

Jakarta, Indonesia · Expected Feb 2027 · GPA 4.00 / 4.00

• Coursework: Business Intelligence & Analytics, Deep Learning, Advanced NLP, Data Warehousing.
• Thesis (in progress, unpublished): a late-fusion transformer framework (IndoBERT, XLM-RoBERTa) with psycholinguistic features to detect workplace stress in Indonesian text.

Jakarta, Indonesia · Completed via fast-track program · GPA 3.78 / 4.00

• Awarded KRW 200,000 for a self-written cultural-exchange grant proposal (Jeju cultural exploration).
• Featured in the BINUS International Office testimonials.

DuckDB · Kimball star schema · scikit-learn · LLM agent · Streamlit

• End-to-end analytics on the Olist e-commerce dataset: a dimensional warehouse, a review-risk model, and a natural-language agent that verifies its own answers (tagging each trusted / repaired / abstained).
• Evaluated on 150 golden questions × 5 models × 3 runs with exact McNemar tests: verification cut confidently-wrong answers by up to 69% (21.8% → 6.7%, p = 1.7e-15); review-risk model reached 0.756 ROC-AUC, flagging the riskiest 10% of orders to catch 41% of all bad reviews. Live demo.

BigQuery · Looker Studio · SQL

• Designed a star-schema dimensional model over the Olist dataset and built an ETL pipeline into BigQuery following Kimball methodology.
• Delivered an interactive Looker Studio dashboard tracking sales, delivery-performance, and customer-satisfaction KPIs for non-technical stakeholders; documented metric definitions and data lineage for governance.

Python · PyTorch · Hugging Face Transformers

• Fine-tuned and benchmarked BERT, RoBERTa, and DistilBERT for binary stress detection on the Reddit subset of the IJCNN 2022 benchmark (“Stress Detection from Social Media Articles”, IEEE WCCI 2022); RoBERTa reached 97.2% accuracy and 0.997 ROC-AUC.
• Vetted the dataset before training and excluded its “non-advertisement” Twitter split, which still contained heavy advertising noise.
• Assessed accuracy-vs-efficiency trade-offs (DistilBERT ~2× faster inference) via macro / weighted F1, per-class metrics, and confusion matrices to guide deployment.

Python · TensorFlow / Keras

• Built and compared three CNN architectures (VGG-19, InceptionV3, ResNet50) with ImageNet transfer learning on CelebA; InceptionV3 led at ~91.5% accuracy.

Python · scikit-learn · LightGBM · XGBoost

• Built an end-to-end classical ML pipeline (TF-IDF features, SMOTE, hyperparameter tuning) comparing seven models on a severely imbalanced ~60k-sample Indonesian dataset.