My Projects

Web Development · 2026

FlowForge: Multi-Tenant Workflow Orchestration Engine

A self-hosted workflow orchestration engine in the style of Zapier/GitHub Actions. Teams define workflows as a DAG of steps (HTTP call, script, delay, conditional branch) triggered manually, on a schedule, or via signed webhook, with real-time run visualization. Multi-tenant isolation is enforced at the database level using PostgreSQL Row-Level Security, with Postgres SKIP LOCKED/LISTEN-NOTIFY as the job queue and Server-Sent Events for realtime updates. Shipped with CI, Docker Compose deployment, and published container images.

PythonFastAPIReact+3

Audio / Bioacoustics · 2026

BirdCLEF+ 2026: Bioacoustic Species Classification

A bird/bioacoustic species classification pipeline built for the Kaggle BirdCLEF+ 2026 competition, covering 234 species. Combines an EfficientNet mel-spectrogram baseline with cached Perch embeddings and a custom state-space model (ProtoSSM) trained with focal BCE and knowledge distillation, ensembled with per-class MLP probes and a residual state-space model, then exported to ONNX for Kaggle inference submission.

PythonPyTorchscikit-learn+1

Information Retrieval · 2026

MC-Retrieval (Mochi): Cross-Modal Retrieval for 3D Minecraft Structures

A cross-modal retrieval system mapping natural language queries and 3D voxel Minecraft structures into a shared semantic embedding space using a CLIP-style dual encoder (CNN and Point-BERT backbones, Masked Voxel Modeling pretraining). Evaluated against TF-IDF, random, and unimodal baselines using Recall@K and MRR; includes an interactive Next.js web demo supporting search by free-text description or by 3D voxel structure built in the browser.

PythonPyTorchNext.js+1

NLP / LLM · 2025

LLaDA: Diffusion Model PEFT

Implementation of Parameter-Efficient Fine-Tuning for diffusion models

PythonPyTorchHuggingface+1

Web Development · 2025

FinSight - Financial Management Platform for UMKM

FinSight is a comprehensive financial management platform designed to help Indonesian UMKM (Usaha Mikro, Kecil, dan Menengah) make better business decisions through smart financial analysis, cash flow predictions, and AI-based business recommendations.

FastAPIHTMLCSS+1

NLP / LLM · 2025

Comparative Performance Analysis of Medical LLMs through QLoRA Adaptation

This project aims to develop a reliable and safe Indonesian-language medical chatbot specifically focused on digestive system health. We analyzed and fine-tuned 7 open-source Large Language Models (LLMs) using QLoRA (Quantized Low-Rank Adaptation) on a dataset of 21,376 doctor-patient Q&A pairs sourced from Alodokter.

PyTorchHuggingfaceQLoRA+1

Time Series / Forecasting · 2025

JPY Currency Exchange Rate Prediction Using RNN and LSTM

This repository implements Recurrent Neural Network (RNN) and Long Short-Term Memory (LSTM) models to forecast the foreign exchange rate of the Japanese Yen (JPY).

PythonPyTorchRNN+1

Computer Vision · 2025

GI-Polyp: Deep Learning Object Detection for Gastrointestinal Polyps

Designed and implemented a gastrointestinal polyp object detection system using deep learning. The project uses a clinically relevant gastroenteroscopic polyp dataset with annotated bounding boxes and benchmarks multiple state-of-the-art detection models (e.g., YOLOv8, YOLOv11, and transformer-based RF-DETR). Evaluated models using standard metrics such as IoU, mAP@50, mAP@50-95, precision, and recall to compare performance in real medical imaging conditions.

PythonPyTorchUltralytics YOLO+1

Agentic AI · 2025

MiningChain: Agent-Based System for Mining Value Chain Optimization

MiningChain is an agent-based decision support system that optimizes mining value chain operations using multi-agent simulation and operational data integration. It provides scenario-based recommendations to improve efficiency, mitigate risk, and support planning decisions across mining, logistics, and shipping activities.

PythonOpenAI Agent SDKMulti-Agent Systems

Biomedical AI · 2025

Adaptive Segment Selection Framework for Transcranial Doppler Signals

Developed an end-to-end framework for automated analysis of Transcranial Doppler (TCD) signals to address data scarcity caused by poor signal quality. Implemented a Signal Quality Assessment (SQI) pipeline to stratify signals, followed by CycleGAN-based signal restoration to recover borderline-quality data. Trained and evaluated multiple 1D deep learning architectures, including CNN, ResNet, LSTM, Transformer, and a custom Attention CNN, for hemodynamic classification.

PythonPyTorchMATLAB+2

NLP / LLM · 2025

Transfer Learning Using LoRA+ on Llama 3.2 for Indonesian Conversations

Applied Parameter-Efficient Finetuning (LoRA+) to adapt Llama 3.2 1B for Indonesian conversational tasks. Achieved a 112% improvement in BLEU and a 21.7% increase in ROUGE-L compared to baseline models, demonstrating effective low-compute cross-lingual transfer learning despite domain shift effects.

PythonPyTorchUnsloth+2