I am a 7th-semester Informatics Engineering student at Universitas Dian Nuswantoro (UDINUS), Indonesia. My journey is driven by curiosity—turning raw data into meaningful stories. I specialize in mining insights from structured numbers (Tabular) and unstructured language (Text/NLP).
- 🧐 What I'm Focusing On
- 🛠️ Tech Stack & Tools
- 🚀 Featured Projects
- ✍️ Latest Blog Posts
- 📈 GitHub Stats
- 📫 Connect with Me & Resume
- Data Science: Exploratory Data Analysis (EDA), Feature Engineering, and Model Building.
- NLP (Natural Language Processing): Text Classification, Sentiment Analysis, and Language Modeling.
- Tabular Data: Handling imbalanced datasets (SMOTE/SMOTENC), preventing data leakage, and interpretability.
Languages:
Data Science & Machine Learning:
NLP, GenAI & LLMOps:
Web Frameworks & UI:
Automation, Bots & DevOps:
Databases:
Data Analytics & Visualization:
Environment & Tools:
-FFCA28?style=for-the-badge&logo=firebase&logoColor=black){kind=icon}
Technologies I am actively learning and integrating into upcoming projects.
Here is a comprehensive list of my Data Science & AI projects, ranging from robust predictive modeling to API deployment and NLP exploration:
🌟 Top Repositories:
Interactive AI agents and chatbots powered by RAG and LLMs.
| Project Name | Domain | Key Tech | Description |
|---|---|---|---|
| Auto-README Generator | GenAI / DevTool | CrewAI, Gemini 2.5 Flash, Streamlit | Agentic Documentation Builder. A multi-agent AI system that deeply analyzes Python/Jupyter codebases and synthesizes them with strategic business metrics. Instantly generates CTO-ready, industry-standard Markdown documentation. |
| InsightSQL (LangGraph Engine) | GenAI / Graph Agent | LangGraph, Gemini 2.5 Flash | State-of-the-Art Reasoning. The evolution of SQL Agents using Graph Architecture. Uses Cyclic Reasoning to query, validate, and self-correct errors in real-time with full "Glass Box" transparency. |
| InsightData (AI Analyst) | GenAI / Data Analyst | Pandas Agent, Gemini 2.5 Flash | Automated Data Scientist. A ReAct agent that analyzes CSVs, Excel, & Google Sheets. Generates Python code for cleaning, deep-dive statistical analysis, and auto-visualization (Matplotlib) on the fly. |
| InsightSQL (ReAct Engine) | GenAI / SQL Agent | LangChain (ReAct), Gemini 2.5 Flash | Advanced Reasoning Agent. Intelligent DBA using ReAct Architecture (Thought-Action-Observation) to reason through messy data. Features Glass Box AI transparency, self-correction, and multilingual business insights. |
| InsightSQL (Simple) | GenAI / SQL Agent | LangChain (SQL Agent), Gemini 2.5 Flash | Zero-Shot SQL Engine. Streamlined Text-to-SQL assistant using standard create_sql_agent. Efficiently maps schema to queries with real-time execution feedback via Streamlit Callbacks. |
| Chat with Vio (Portfolio AI) | GenAI / Portfolio | Streamlit, LangChain, Gemini 2.5 | Interactive CV. A personalized RAG assistant that "chats" with a professional profile. Features Strict Anti-Hallucination, bilingual support (EN/ID), and a responsive UI optimized for recruiters. |
| DocuTalk-AI (Smart RAG) | GenAI / RAG | LangChain, Gemini 2.5 Flash, FAISS | Agentic Workflow. Intelligent document assistant using ReAct Agents to decide between Document retrieval (PDF, CSV, TXT, MD) (FAISS) or Web Search. Features Incremental Indexing (cost-efficient), strict language control, and session export. |
| DocuFlow-AI (Fast RAG) | GenAI / RAG | LangChain, Gemini 2.5 Flash, FAISS | High-Performance Engine. Streamlined RAG using ConversationalRetrievalChain for instant, precise analysis (PDF, CSV, MD). Features Transparent Verification (Source & Score), incremental indexing, and zero agentic latency. |
| DocuChat AI (Long Context) | GenAI / LLM | Gemini 2.5 Flash, Streamlit, Python | Full Context Approach. Analyzing entire PDFs without chunking (RAG alternative) using Gemini's massive context window. Features Strict Language Enforcement and real-time token monitoring. |
| AI Recipe Generator | GenAI / Agent | LangChain, Gemini, Streamlit | Agentic App. Orchestrated a cooking assistant using LangChain ReAct Agents and Gemini 2.5 Flash to detect ingredients (Vision) and generate recipes with conversation memory. |
High-performance APIs and production-ready inference engines.
| Project Name | Domain | Key Tech | Description |
|---|---|---|---|
| Maternal Health Risk API | API / MLOps | FastAPI, Pydantic, CatBoost, LIME | Explainable Risk-Triage Microservice. A RESTful backend serving a tuned CatBoost classifier via /predict and /explain endpoints, returning both a risk classification and an interactive LIME explanation. Features strict Pydantic input validation against clinical vital-sign bounds and structured, multi-category error handling. |
| Resume Scanner API | NLP / ATS | FastAPI, SBERT, TF-IDF | Dual-Engine ATS Optimizer. A stateless API analyzing resumes against job descriptions. Features Strict Mode (TF-IDF) for keyword matching and Flexible Mode (SBERT) for semantic relevance. |
| Portfolio API (Backend) | Backend / API | FastAPI, Docker, FAISS | Headless RAG Engine. The high-performance backend serving the portfolio data. Features Swagger UI for testing, structured JSON responses with Source Citations, and fully Dockerized deployment on Hugging Face Spaces. |
| Stuntify API (MLOps) | API / MLOps | FastAPI, Pydantic, Scikit-Learn | High-Performance Engine. A robust inference system using FastAPI to serve real-time stunting predictions. Features strict Type Coercion, multi-artifact orchestration, and interactive Swagger documentation. (Note: Trained on synthetic data and for demonstration purposes only). |
| Insightify API (NLP) | NLP / API | FastAPI, Transformers, Pandas | Dual-Lingual Analytics. NLP microservice for English & Indonesian sentiment using RoBERTa. Features Batch Processing (Excel/CSV), N-Gram keyword extraction, and automated text statistics. |
| Stunting Prediction API | API / MLOps | Flask, Python | Deployment. Serving the stunting prediction model as a RESTful API for real-time inference using Flask. |
Scheduled workflows, autonomous bots, and task management systems.
| Project Name | Domain | Key Tech | Description |
|---|---|---|---|
| NovaCal AI (Stateful Telegram) | GenAI / Assistant | LangChain, SQL, Gemini Flash | Stateful Calendar Assistant. An advanced Telegram bot featuring an SQL-backed conversational memory architecture. Enables natural, multi-turn dialogue for Google Calendar management (CRUD) without losing context between messages. |
| NovaCal AI (Ephemeral Telegram) | GenAI / Assistant | LangChain, RAM, Gemini Flash | Slot-Filling Calendar Assistant. An optimized Telegram bot utilizing an ephemeral, RAM-based conversational memory. It retains context just long enough to gather missing parameters for a calendar task, then automatically purges its memory upon completion to save API tokens and prevent hallucinations. The ideal middle-ground architecture. |
| NovaCal AI (Stateless Telegram) | GenAI / Assistant | LangChain, PTB, Gemini Flash | Zero-Memory Calendar Assistant. A lightning-fast, stateless Telegram bot utilizing LangChain and Gemini to manage Google Calendar events via strict tool-calling. Optimized for quick, single-turn executions with maximum token efficiency. |
| NovaCal AI (Streamlit Edition) | AI Dashboard | Streamlit, LangChain, Gemini | Interactive AI Calendar Dashboard. A web-based visual interface for intelligent calendar management. Features a stateful LangChain agent utilizing a session-based conversational memory buffer to handle complex scheduling requests alongside a dynamic UI. |
| NovaTasks AI (Ephemeral Telegram) | GenAI / Assistant | LangChain, RAM, Gemini Flash | Slot-Filling Tasks Assistant. A conversational Telegram bot integrated with Google Tasks. Built using LangChain and Google Gemini, it utilizes an ephemeral RAM-based memory architecture to intuitively handle complex, multi-turn to-do list management (CRUD) while aggressively saving API tokens and preventing AI hallucinations. |
| NovaMail AI (Gmail Autoresponder) | GenAI / Email | Gmail API, Gemini Flash, Requests | Autonomous Inbox Assistant. An automated bot that continuously checks your Gmail for new messages. It uses Gemini AI to read the emails and automatically send smart, professional replies. It keeps the replies neatly organized in the original email thread and instantly sends you a summary report on Telegram. This is a general-purpose implementation; for further customization, please refer to the repository README. |
| Daily Agenda Broadcaster | Automation | Python, Google Calendar API | Scheduled Notification Engine. An automated workflow that triggers at designated intervals to fetch the day's upcoming events directly from Google Calendar and dispatches a clean, formatted briefing to a designated Telegram chat. |
| AI Vision Generator | GenAI / Vision | PTB, Hugging Face, SDXL | Text-to-Image Telegram Bot. Integrates with the Hugging Face Inference API (stabilityai/stable-diffusion-xl-base-1.0) to instantly render high-quality images based on user text prompts and seamlessly deliver them directly back to the Telegram chat. |
| Conversational AI Interface | GenAI / Chatbot | PTB, Gemini Flash | Interactive LLM Bridge. A streamlined integration bridging Telegram and Google's Gemini model using the python-telegram-bot framework. Processes user prompts in real-time and delivers intelligent, generative responses natively within the chat application. Excellent as a foundational learning resource for python-telegram-bot (PTB) beginners, demonstrating a clear, end-to-end data flow: capturing user input from Telegram, processing it through the Gemini API, and delivering the generated response directly back to the chat interface. |
| Daily AI News Digest | Orchestration | Prefect, DuckDuckGo, Gemini Flash | Autonomous Intelligence Agent. A fully automated pipeline that scrapes real-time news via DuckDuckGo, synthesizes summaries with Gemini, and delivers daily digests to Telegram with fail-fast logic. |
| Automated Image Pipeline | Orchestration | Prefect, Hugging Face, GSheets | Consumer Bot (Art Gen). The downstream consumer in the Creative Ecosystem. It reads prompts from GSheets (fed by either the Daily Prompt Generator or Batch Prompt Generator), generates AI art via Hugging Face, and publishes to Telegram. |
| Daily Prompt Generator | GenAI / ETL | Prefect, Gemini Flash, GSheets | Producer Bot (1 Item). Wakes up daily to craft exactly 1 prompt using Gemini and logs it to GSheets. Generates 1 prompt per execution. Designed to work with the Automated Image Pipeline. See Batch Version if you want to generate 3 prompts per run. |
| Batch Prompt Generator | GenAI / ETL | Prefect, Gemini Flash, GSheets | Producer Bot (3 Items). Wakes up daily to craft exactly 3 prompts in a single API call and bulk-uploads them to GSheets. Efficient for filling the queue fast. Designed to work with the Automated Image Pipeline. See Standard Version if you only want 1 prompt per run. |
| Daily Quote Automator (Standard Version) | Automation | Prefect, Gemini, Requests | Simple HTTP Bot. A lightweight automation using direct requests (Synchronous). Wakes up daily to synthesize developer motivation. See PTB Version for the Library implementation. |
| Daily Quote Bot (PTB Version) | Automation | Prefect, Gemini, Async | Library Wrapper Bot. An alternative implementation using the official python-telegram-bot library. See Standard Version if you prefer the lighter requests method. |
In-depth data analysis, model training, NLP implementation, and academic research.
| Project Name | Domain | Key Tech | Description |
|---|---|---|---|
| Maternal Health Risk (Research) | Tabular | CatBoost, imbalanced-learn, LIME | Clinical Risk Benchmarking. Compared 6 classifiers against 6 class-imbalance handling techniques to predict maternal health risk from vital signs, then tuned and interpreted the best model with LIME. (Trained on the UCI Maternal Health Risk dataset from rural Bangladesh). |
| Maternal Health Risk (Streamlit) | Tabular / App | Streamlit, CatBoost, LIME | Dual-Architecture Demo App. Two parallel Streamlit dashboards for the same risk model: a self-contained version running inference in-process, and a client version consuming a separate FastAPI backend, illustrating monolithic vs. decoupled deployment patterns side by side. |
| Maternal Health Risk (Gradio) | Tabular / App | Gradio, CatBoost, LIME | Dual-Architecture Demo App. The Gradio counterpart to the Streamlit app above, offering the same in-process vs. API-client comparison with a Gradio-based UI and embedded LIME explanations. |
| Depression Predictor | Tabular | Scikit-Learn, Streamlit, ImbPipeline | Comparative ML Pipeline. An end-to-end web app demonstrating a complete workflow: baseline modeling, handling rare categories, and addressing class imbalance. (Note: Developed for a university assignment to demonstrate preprocessing techniques). |
| SpendSense (Financial AI) | GenAI / App | Streamlit, LangChain, Gemini Vision | Conversational Finance App. Integrates Pandas Agent for natural language queries ("How much did I spend on food?"), Gemini Vision for OCR receipt scanning, and automated ETL with Google Sheets. |
| Streamlit Resume Scanner | NLP / Applied ML | SBERT, Scikit-Learn, Streamlit | Dual-Engine CV Analyzer. A hybrid NLP tool combining Strict ATS Logic (TF-IDF) and Semantic Search (SBERT Embeddings) to calculate resume-job fit and detect missing keywords. |
| Diabetes Prediction (Robust) | Tabular | Scikit-Learn, SMOTE | Best Practice. Built a robust model ensuring no data leakage by applying Scaling & SMOTE after data splitting. |
| Stunting Analysis | Tabular | SMOTENC, Tree Viz | Handling categorical imbalance with SMOTENC and visualizing decision paths (Pruned Tree) for interpretability. (Note: Trained on synthetic data and for demonstration purposes only). |
| Review Sentiment Analyzer | NLP | Transformers, Python | Leveraging Hugging Face pre-trained models to accurately classify text sentiment (Positive/Negative). |
| Diabetes Paper Replication | Academic | Python, Pandas | Research Study. Replicating a specific research paper's methodology (including its pre-split scaling technique) for comparative analysis. |
Focusing on exploratory data analysis (EDA), business intelligence dashboards, and visual storytelling.
| Project Name | Domain | Key Tech | Description |
|---|---|---|---|
| Supermarket Sales Dashboard | Retail Analytics | Tableau Public | Sales & Operations. Dashboard monitoring Total Sales, COGS, and Customer Ratings across major cities. Features breakdown by Payment Methods (E-wallet/Cash) and Product Line performance. |
| Bike Sales Dashboard | Executive Sales | Tableau Public | Revenue Tracker. High-level executive view of Revenue and Unit Sales. Analyzes performance by US States, Brand Names, and identifies Top 10 High-Value Customers. |
Creative coding projects exploring logic flows, real-time communication, and game-like interactions.
| Project Name | Domain | Key Tech | Description |
|---|---|---|---|
| Conversational AI Agent | GenAI / Persona | Streamlit, Gemini LLM | Roleplay Chatbot. An engaging AI companion simulating a romantic partner using Google Gemini, featuring persistent memory and dynamic persona adaptation. |
| Real-time Chat Group | Social / Socket | Python, Networking | Multi-User Chat. A casual project exploring real-time communication and socket programming principles to create a functional group chat environment. |
⚠️ Disclaimer: These projects are created purely for educational purposes and fun. They are provided "as-is" without warranty of any kind. The author is not responsible for any security vulnerabilities, data loss, or misuse resulting from the deployment of this code. Not intended for production use.
📝 Before You Use AI, Understand This First(Published: May 29, 2026)
📝 Why Label Encoding is Ruining Your ML Models (And How to Fix It)(Published: Apr 19, 2026)
🐱 My GitHub Data
📦 187.5 kB Used in GitHub's Storage
🏆 860 Contributions in the Year 2026
💼 Opted to Hire
📜 78 Public Repositories
🔑 10 Private Repositories
I'm an Early 🐤
🌞 Morning 346 commits ███████░░░░░░░░░░░░░░░░░░ 29.35 % 🌆 Daytime 680 commits ██████████████░░░░░░░░░░░ 57.68 % 🌃 Evening 152 commits ███░░░░░░░░░░░░░░░░░░░░░░ 12.89 % 🌙 Night 1 commits ░░░░░░░░░░░░░░░░░░░░░░░░░ 00.08 % 📅 I'm Most Productive on Friday
Monday 125 commits ███░░░░░░░░░░░░░░░░░░░░░░ 10.60 % Tuesday 159 commits ███░░░░░░░░░░░░░░░░░░░░░░ 13.49 % Wednesday 167 commits ████░░░░░░░░░░░░░░░░░░░░░ 14.16 % Thursday 147 commits ███░░░░░░░░░░░░░░░░░░░░░░ 12.47 % Friday 257 commits █████░░░░░░░░░░░░░░░░░░░░ 21.80 % Saturday 197 commits ████░░░░░░░░░░░░░░░░░░░░░ 16.71 % Sunday 127 commits ███░░░░░░░░░░░░░░░░░░░░░░ 10.77 % 📊 This Week I Spent My Time On
🕑︎ Time Zone: Asia/Jakarta
💬 Programming Languages: Python 12 hrs ███████████████████████░░ 92.51 % Bash 43 mins █░░░░░░░░░░░░░░░░░░░░░░░░ 05.58 % Nix 7 mins ░░░░░░░░░░░░░░░░░░░░░░░░░ 00.96 % Text 3 mins ░░░░░░░░░░░░░░░░░░░░░░░░░ 00.49 % Other 1 min ░░░░░░░░░░░░░░░░░░░░░░░░░ 00.19 % 🔥 Editors: VS Code 12 hrs 58 mins █████████████████████████ 100.00 % 🐱💻 Projects: ai-resep 12 hrs 58 mins █████████████████████████ 100.00 % 💻 Operating System: Linux 12 hrs 58 mins █████████████████████████ 100.00 % 🤖 AI Coding This Week
No AI Coding Activity Tracked This Week
I Mostly Code in Python
Python 46 repos ███████████████░░░░░░░░░░ 58.23 % Jupyter Notebook 13 repos ████░░░░░░░░░░░░░░░░░░░░░ 16.46 % Swift 1 repo ░░░░░░░░░░░░░░░░░░░░░░░░░ 01.27 % TypeScript 1 repo ░░░░░░░░░░░░░░░░░░░░░░░░░ 01.27 % JavaScript 1 repo ░░░░░░░░░░░░░░░░░░░░░░░░░ 01.27 % Last Updated on 10/08/2026 01:05:13 UTC
-ED2224?style=for-the-badge&logo=adobe-acrobat-reader&logoColor=white){kind=icon}
💜 If you find my work interesting, consider giving a star to my repositories! 💜
"Data is the new oil, but it's useless without the engine to refine it."