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 |
|---|---|---|---|
| YouTube Vector RAG Chatbot | GenAI / Vector RAG | LlamaIndex, ChromaDB Cloud, Gemini, LiteLLM | Chat with a YouTube Video's Transcript. A CLI RAG chatbot with a self-written transcript loader that embeds the captions in ChromaDB Cloud and answers strictly from the retrieved passages, using cross-encoder reranking and persistent PostgreSQL memory. Features Ingest-Once indexing, Arize Phoenix tracing, per-turn token and cost reporting, and the source nodes printed behind every answer. (The Vector baseline of a three-architecture series. See the Graph version and the Hybrid version). |
| YouTube Graph RAG Chatbot | GenAI / Graph RAG | LlamaIndex, Neo4j, Gemini, LiteLLM | Chat with a YouTube Video Through a Knowledge Graph. The Graph version of the series: Gemini extracts relationships from every transcript chunk into a Neo4j property graph, and each question retrieves the related parts of the graph together with their source text, then reranks them with a cross-encoder before answering strictly from them. Features Build-Once graph storage, persistent PostgreSQL memory, Arize Phoenix tracing, and per-turn token and cost reporting. (See the Vector version and the Hybrid version). |
| YouTube Hybrid RAG Chatbot | GenAI / Hybrid RAG | LlamaIndex, ChromaDB Cloud, Neo4j, Gemini, LiteLLM | Vector Search and Knowledge Graph, Fused. The Hybrid version of the series: the LLM writes extra variants of each question, every variant is searched in both ChromaDB Cloud and a Neo4j graph, and the results are merged with Reciprocal Rank Fusion before cross-encoder reranking. Features persistent PostgreSQL memory, Arize Phoenix tracing, and per-turn token and cost reporting. (See the Vector version and the Graph version, and the benchmark that compares all three). |
| NL2SQL (All Schema Context) | GenAI / Text-to-SQL | LlamaIndex, LiteLLM, Gemini, PostgreSQL | Full-Schema Text-to-SQL. Two CLI scripts, single-turn and multi-turn, that give Gemini the schema of every database table through LlamaIndex's NLSQLTableQueryEngine, with no vector database and no embedding model. Hand-written table descriptions are forced into the prompt so the generated SQL respects exact column spellings and value formats. The chat version stores its history in PostgreSQL to handle follow-up questions. Features LiteLLM model fallback, Arize Phoenix tracing, per-question token and cost reporting, and runs on PostgreSQL or the included SQLite sample. (See the Selected Schema version for the retrieval-based alternative). |
| NL2SQL (Selected Schema Retrieval) | GenAI / Text-to-SQL | LlamaIndex, ChromaDB Cloud, Gemini, PostgreSQL | Schema-Retrieval Text-to-SQL. Two CLI scripts, single-turn and multi-turn, that embed each table's hand-written description in ChromaDB Cloud and retrieve only the top 3 closest schemas per question, so the LLM writes SQL from just the tables it needs. The chat version keeps its history in PostgreSQL to handle follow-up questions. Features Index-Once Ingestion, LiteLLM model fallback, Arize Phoenix tracing, per-question token and cost reporting, and the generated SQL with its source rows printed behind every answer. Includes a SQLite sample database. (See the All Schema version for the variant without a vector database). |
| LlamaIndex RAG & Agents Practice | GenAI / RAG Practice | LlamaIndex, ChromaDB, Gemini, Cross-Encoder | Step-by-Step LlamaIndex Practice. Six progressive CLI scripts, from a basic RAG query engine, through chunking, saved indexes, and local ChromaDB, to a context chat engine with memory and finally FunctionAgent and ReActAgent over two reranked document collections. Includes a limitations analysis (RAM and file-based storage) plus notes and code for moving to server-based storage. |
| Website Reader RAG Chatbot | GenAI / RAG | LlamaIndex, ChromaDB Cloud, Gemini, Trafilatura | Web Page Q&A Engine. A CLI RAG chatbot that scrapes web pages with Trafilatura, indexes them in ChromaDB Cloud, and answers strictly from the page content using cross-encoder reranking and persistent PostgreSQL memory. Features Idempotent Ingestion (scrape and embed once, reuse on later runs) and categorized error handling. |
| YouTube Multimodal RAG Chatbot | GenAI / Multimodal RAG | LlamaIndex, CLIP, ChromaDB Cloud, Gemini 2.5 Flash | Chat with a Video's Transcript and Frames. A CLI assistant that downloads a YouTube video, then indexes its transcript (Gemini embeddings) and sampled frames (CLIP) as separate ChromaDB collections and answers from both. Features cross-encoder reranking, persistent PostgreSQL memory, Arize Phoenix tracing, per-turn token/cost reporting, and the exact source nodes printed behind every answer. |
| RAG Token Benchmark (Vector vs Graph vs Hybrid) | GenAI / RAG Benchmark | LlamaIndex, ChromaDB, Neo4j, LiteLLM | Cost-per-Architecture Study. Benchmarks token consumption and estimated USD cost of Vector, Graph, and Hybrid RAG on the same YouTube transcript and the same three questions (simple, analytical, multi-hop), with token counting, step-level tracing, and LiteLLM cost estimation. Result: cost rose from Vector to Graph to Hybrid on every question. |
| 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. Live Demo |
| 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. Live Demo |
| 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. Live Demo |
| 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. Live Demo |
| 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. Live Demo |
| 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. Live Demo |
| 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. Live Demo |
| 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. Live Demo |
| 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. Live Demo |
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. Live Demo |
| 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. Live Demo |
| 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). Live Demo |
| 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. Live Demo |
| Stunting Prediction API | API / MLOps | Flask, Python | Deployment. Serving the stunting prediction model as a RESTful API for real-time inference using Flask. Live Demo |
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. Live Demo |
| 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 |
|---|---|---|---|
| Medical Cost Prediction | Tabular / Regression | Random Forest, LightGBM, Optuna, LIME | 9-Model Regression Benchmark. Compared 9 regressors to predict medical insurance charges, tuned the top two (Random Forest, LightGBM) across GridSearchCV, RandomizedSearchCV, and Optuna, then added LightGBM quantile regression for prediction intervals and LIME for per-patient explainability. Kaggle Notebook |
| Movie Recommendation (SVD) | Tabular / Recommender | scikit-surprise, GridSearchCV | Collaborative Filtering Engine. Built an SVD-based movie recommender on the MovieLens dataset, covering duplicate-ID data cleaning, EDA, 5-fold cross-validation, and GridSearchCV hyperparameter tuning to generate personalized top-N recommendations. (See the ALS companion project below for a distributed alternative on the same problem). Kaggle Notebook |
| Movie Recommendation (ALS / PySpark) | Tabular / Distributed | PySpark MLlib, ALS, CrossValidator | Distributed Collaborative Filtering. The Spark-based companion to the SVD recommender above, solving the same MovieLens recommendation problem with Alternating Least Squares (ALS) matrix factorization, built to scale beyond what a single-machine library can handle. Kaggle Notebook |
| 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. |
| Maternal Health Risk (Binary + LIME) | Tabular / App | Streamlit, AdaBoost, LIME | Single-Patient Risk Triage App. A monolithic Streamlit dashboard predicting a binary (Low Risk / Elevated Risk) maternal health classification from six vitals, with an instant LIME explanation rendered alongside every prediction. Live Demo |
| 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). Live Demo |
| 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. Live Demo |
| 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. Live Demo |
| 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. |
Interactive web apps, from front-end-only experiences to full-stack platforms, built end-to-end through "vibe coding" (AI-assisted, prompt-driven development in Google AI Studio with Gemini), showcasing versatility beyond my core Data Science/AI focus.
| Project Name | Domain | Key Tech | Description |
|---|---|---|---|
| ForeverCard (Anniversary Card) | Vibe Coding / Web App | React, TypeScript, Vite, Web Audio API | Interactive Celebration Card. A romantic 1-year anniversary web card featuring a cursor-dodging "no" button, a polaroid relationship-timeline gallery, live selfie capture with filters, a love quiz, a downloadable keepsake letter, confetti bursts, and synthesized background music. Live Demo |
| WishBox (Birthday Card) | Vibe Coding / Web App | React, TypeScript, Vite, Web Speech API | Interactive Celebration Card. A festive, relationship-agnostic birthday web card guiding the recipient through making a wish (real microphone candle-blow detection), unwrapping a gift box, and reading a personal letter with confetti, synthesized music, and browser text-to-speech. Live Demo |
| LinkNest (Link-in-Bio Platform) | Vibe Coding / Full-Stack App | React, TypeScript, Supabase, Gemini | Full-Stack Link-in-Bio Platform. A customizable link-in-bio platform with a public profile page and a Supabase-backed admin dashboard, featuring AI-generated link descriptions, a Google Search powered Discover tab, broken-link checks, password-protected pages, dynamic Open Graph previews, and click analytics. Live Demo |
| AutoVista Motors (Car Dealership Showcase) | Vibe Coding / Web App | React, TypeScript, Vite, Tailwind CSS | Multipage Car Dealership Showcase. A fictional 7-page dealership site with a 24-vehicle lineup, featuring live search, sorting, budget filtering, side-by-side vehicle comparison, a working financing calculator, and a multi-layer parallax homepage, built as a front-end design showcase. Live Demo |
| Hanami (Japan Destination Showcase) | Vibe Coding / Web App | React, TypeScript, Tailwind CSS, Leaflet | Japan Travel Journal Showcase. A fictional 7-page Japan travel journal with 16 destinations, four seasonal themes that restyle the whole site, an interactive regional map, a drag-and-drop itinerary builder, six sample travel packages, and a searchable travel tips guide and FAQ, built as a front-end design showcase. Live Demo |
| Skycast (Weather Dashboard) | Vibe Coding / Web App | React, TypeScript, Vite, Open-Meteo API | Live Weather Dashboard. A weather dashboard with hourly and 7-day forecasts, air quality, rule-based alerts and outfit tips, a live rain radar, side-by-side city comparison, a historical climate view, and ambient sounds that match the weather, generated live with the Web Audio API. No backend and no API key needed. Live Demo |
| Botanica AI (Plant Care App) | Vibe Coding / Full-Stack App | React, TypeScript, Express, Gemini | Full-Stack Plant Care Companion. A React web app with an Express API proxy that identifies a plant from a photo, diagnoses leaf health, and generates an 8-point care plan, featuring a My Garden watering tracker, Dr. Flora, a multi-turn AI botanist chat, a potting-soil calculator, structured JSON output, cascaded Gemini model fallback, and rate-limited server-side API calls. Live Demo |
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
📦 280.0 kB Used in GitHub's Storage
🏆 998 Contributions in the Year 2026
💼 Opted to Hire
📜 92 Public Repositories
🔑 12 Private Repositories
I'm an Early 🐤
🌞 Morning 420 commits ████████░░░░░░░░░░░░░░░░░ 31.91 %
🌆 Daytime 735 commits ██████████████░░░░░░░░░░░ 55.85 %
🌃 Evening 160 commits ███░░░░░░░░░░░░░░░░░░░░░░ 12.16 %
🌙 Night 1 commits ░░░░░░░░░░░░░░░░░░░░░░░░░ 00.08 %
📅 I'm Most Productive on Friday
Monday 132 commits ███░░░░░░░░░░░░░░░░░░░░░░ 10.03 %
Tuesday 171 commits ███░░░░░░░░░░░░░░░░░░░░░░ 12.99 %
Wednesday 190 commits ████░░░░░░░░░░░░░░░░░░░░░ 14.44 %
Thursday 171 commits ███░░░░░░░░░░░░░░░░░░░░░░ 12.99 %
Friday 284 commits █████░░░░░░░░░░░░░░░░░░░░ 21.58 %
Saturday 222 commits ████░░░░░░░░░░░░░░░░░░░░░ 16.87 %
Sunday 146 commits ███░░░░░░░░░░░░░░░░░░░░░░ 11.09 %
📊 This Week I Spent My Time On
🕑︎ Time Zone: Asia/Jakarta
💬 Programming Languages:
Python 3 hrs 45 mins ████████████████████████░ 95.93 %
Bash 9 mins █░░░░░░░░░░░░░░░░░░░░░░░░ 04.07 %
🔥 Editors:
VS Code 3 hrs 54 mins █████████████████████████ 100.00 %
🐱💻 Projects:
ai-resep 3 hrs 54 mins █████████████████████████ 100.00 %
💻 Operating System:
Linux 3 hrs 54 mins █████████████████████████ 100.00 %
🤖 AI Coding This Week
No AI Coding Activity Tracked This Week
I Mostly Code in Python
Python 50 repos █████████████░░░░░░░░░░░░ 53.19 %
Jupyter Notebook 16 repos ████░░░░░░░░░░░░░░░░░░░░░ 17.02 %
HTML 12 repos ███░░░░░░░░░░░░░░░░░░░░░░ 12.77 %
TypeScript 8 repos ██░░░░░░░░░░░░░░░░░░░░░░░ 08.51 %
Swift 1 repo ░░░░░░░░░░░░░░░░░░░░░░░░░ 01.06 %
Last Updated on 01/11/2026 03:28:47 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."