Simple AI Chat with Google Generative Language API (Gemini API)
2024
About This Project
Arifian.AI v2.0
Major upgrade of the Arifian.AI personal chatbot. Rebuilt from the ground up with a modern RAG (Retrieval-Augmented Generation) architecture. Now powered by GPT-OSS 120B via Groq for fast, high-quality responses, with multilingual SentenceTransformer embeddings (paraphrase-multilingual-MiniLM-L12-v2) supporting 50+ languages including Indonesian. Features include a dedicated admin dashboard for knowledge base management, system prompt customization, file uploads, RAG configuration, and user management. Backend built with FastAPI and MongoDB Atlas vector search, hosted on Hugging Face Spaces with Docker.
- Python
- FastAPI
- MongoDB
- React
- TypeScript
- Tailwind CSS
Elara : Personal Assistant
Autonomous AI Personal Assistant and RAG engineering engine built with FastAPI (Python 3.11+) and a Hybrid RAG Pipeline fusing dense Gemini Embedding 2 (768d) with Supabase PostgreSQL pgvector 0.8.2 and Full-Text Search (FTS GIN) via Reciprocal Rank Fusion (RRF), reranked by Gemma 4 26B (`gemma-4-26b-a4b-it`) and generated by Groq GPT-OSS 120B. Features a 6-step conversational state machine for project intake and lead qualification with automated persistence to Supabase and real-time outbound alerts to Arifian's Telegram DM / Hermes Agent, along with a built-in Single Page Admin Console (Minimalist Brutalism) for knowledge base management, GitHub repo ingestion, dynamic system prompts, and latency monitoring.
- Python
- FastAPI
- Google Gemini Embeddings & Gemma
- Supabase
- PostgreSQL pgvector
- Cloudflare R2
- Telegram Bot API
- Next.js
Rainfall Prediction with LSTM + Attention
This project explores the use of LSTM (Long Short-Term Memory) and LSTM with an Attention Layer to predict daily rainfall rates in Batam City. The primary goal is to assess the performance difference between the two models when the Attention Layer is implemented. The study leverages the BMKG DataOnline climate dataset and aims to improve prediction accuracy, especially in the context of dynamic and extreme weather changes.
- Python
- Jupyter
- TensorFlow
- Keras
- Scikit-learn
- Flask