Portfolio


Showcasing My Skills & Projects

About Me

Joshua Daniel Talahatu · AI Engineer & Backend Developer

I am an Informatics Engineering student at the University of Surabaya, passionate about building backend systems and implementing machine learning models. My focus areas include:

  • 📚 Learning and applying NLP & Text Classification
  • ⚙️ Developing RESTful APIs with Python (Django, Flask)
  • 🤖 Automating CI/CD with GitHub Actions, Selenium, Puppeteer

Core Skills

Kategori Tools & Teknologi
Languages & Frameworks Python, Django, Flask, Hapi.js, .NET, PHP
Machine Learning TensorFlow, Keras, Scikit‑learn, NLTK, spacy
CI/CD & Automation GitHub Actions, Jenkins, Selenium, Puppeteer
Backend & Infrastructure DBeaver, PostgreSQL, MySQL, Redis, AWS, Docker
Collaboration & Others Git, GitHub, Slack, Discord, Postman

1. AIRA (ML Piano Performance Analyzer)

A backend and ML service that analyzes user-uploaded MIDI files to classify four types of piano performance mistakes by comparing against ideal reference tracks. Custom synthetic dataset is generated based on “Twinkle Twinkle Little Star” melody to train and simulate piano mistake recognition. Methods are inspired outlined in the paper and are applied to identify and correct musical mistakes.

  • Tech Stack: Tensorflow, Flask Framework, Firebase, GCP
  • Highlight: Dataset generation, MIDI conversion, audio feature extraction, CI/CD
  • Org GitHub Repo
  • ML GitHub Repo

2. Version Control via AST

An open-source tool designed to automatically trace and visualize Abstract Syntax Trees (AST) from Python scripts. This enables users to analyze, compare, and understand the transformation of code during development or experimentation without altering the original script.

3. Drone vs Bird Classification

A classification model that distinguishes images of drones and birds using data augmentation, HOG feature extraction, and multiple ML classifiers.

  • Tech Stack: Python, Flask, OpenCV, Seaborn, PHP, Docker Compose
  • Highlight: Model evaluation, HOG feature extraction, hyperparameter tuning with GridSearchCV
  • GitHub Repo

4. Social Media Content Search & Filtering

A service that scrapes social media platforms using Selenium to collect content and filters it based on keyword proximity to deliver targeted insights.

  • Tech Stack: Python, Selenium, PHP
  • Highlight: Automated web scraping, dynamic content handling, keyword-based filtering, content similarity
  • GitHub Repo

Certifications