Web, Android & Odoo.
Interfaces, data & the logic between.

A personal collection of
web, workflows & experiments.

Explore my work

I connect interfaces to the logic behind them.
From offline recipes to shared video and everyday workflows.

Built around everyday questions.

What can I cook? Who borrowed this?
Can we watch together? What needs doing today?

More explorations
Selected work
1 / 4
Continue to About

Android / On-device ML

SmartPantry

From a photo.To a recipe.

Take a photo, review the ingredients, and find recipes. The main flow works on the device, without a server.

GitHub (opens in a new tab)

Full-stack / Laravel

Inventory Management

Borrow it.Track it.Return it.

A catalog and borrowing workflow that keeps requests, approvals, returns, and item availability connected.

GitHub (opens in a new tab)

Real-time / React

Watch Party

One room.One timeline.

Start a room, choose a video, and let connected viewers follow the host playback timeline.

GitHub (opens in a new tab)

Local-first / JavaScript

Local To-Do List

Plan the day.Keep the record.

Tasks, deadlines, attachments, and a flexible logbook in a browser application that stores its data locally.

GitHub (opens in a new tab)

Still curious.

TomatoAI Image / Machine learning

A TensorFlow image classification experiment with a Streamlit interface for inspecting tomato condition predictions.

View on GitHub (opens in a new tab)
Bird song classification Audio / Machine learning

Exploring audio classification and prediction workflows with Python.

View on GitHub (opens in a new tab)
Virtual gifts Laravel / Web

A Laravel project exploring virtual gift interactions with a Vite-powered frontend.

View on GitHub (opens in a new tab)
Animal classification Neural networks / Exploration

An image classification experiment with an artificial neural network and a Gradio interface.

View on GitHub (opens in a new tab)

I like figuring out
how things fit together.

I'm Muhammad Ferry Saputra, a developer working across web, Android, and Odoo. My projects connect what people see on screen to what happens behind it: permissions, application state, local data, and real-time events.

IT Staff Intern at WABE

At WABE, I help translate inventory and purchasing needs into Odoo workflows and reporting. The work involves understanding existing processes, connecting records, handling permissions, and making documents easier to review.

Explore my skills

Away from the keyboard: gym, gaming, and finding the next thing to learn.

Leave room for correction.

SmartPantry lets people edit detected ingredients before searching for recipes. A prediction is a starting point, and the person using it gets the final say.

Make state changes explicit.

In Inventory Management, approval reduces availability and a return restores it. Connecting those actions to request states keeps the workflow understandable.

Think about where data lives.

Local To-Do List separates text records from media storage and provides backup and restore. Keeping data in the browser also means giving people a way to move it.

The idea comes first.
The tools follow.

Some of the ways I turn
a question into something tangible.

Web experiences

From the interface to the application logic behind it. I build with React and Laravel, and explore real-time interactions.

JavaScript, TypeScript, React, Astro, Laravel, PHP, MySQL, Node.js, Socket.io

Business workflows

Turning operational needs into Odoo workflows, permissions, confirmation steps, and documents people can actually use.

Odoo 16, Python, ORM, XML views, PostgreSQL, QWeb, XlsxWriter, Git

Apps & experiments

On-device ingredient detection, local recipe matching, and academic experiments in image and audio classification.

Kotlin, Jetpack Compose, CameraX, LiteRT, Room, TensorFlow, scikit-learn, Streamlit

Have something worth figuring out?
I'd like to hear about it.

Open to roles, projects & collaborations.

LET'S TALK

Tell me what you're building,
where you need help, or what caught your eye.

Android / On-device ML

SmartPantry

Connected photo capture, ingredient detection, editable ingredient selection, and local recipe matching in an Android app.

View source on GitHub (opens in a new tab)

The problem

What can I cook with the ingredients I already have? A photo can help identify them, but people still need a way to correct the selection.

What it enables

Users can take or select a photo, adjust the ingredients, and open recipe ingredients and cooking steps from the local database.

Keep detection and recipes on the device.

A bundled model and a Room database let the main flow work offline. Ingredient selection stays editable so the model prediction can be reviewed before searching.

How I put it together

  • Integrated CameraX and photo selection with a LiteRT ingredient detector.
  • Added ingredient selection and recipe ranking based on the ingredients available.
  • Separated app, data, and domain code, with recipe details loaded from bundled SQLite data.

Tools behind it

Kotlin, Jetpack Compose, CameraX, LiteRT, YOLO11n, Room, SQLite

Offline emulator workflow documented in the repository. Detection accuracy on real ingredient photos and physical-device performance remain separate validation work. Read the validation notes (opens in a new tab).

Full-stack / Laravel

Inventory Management

Built role-based catalog and request flows, with approval and return actions tied to inventory availability.

View source on GitHub (opens in a new tab)

The problem

A borrowing request should not reserve stock before it is approved, and returning an item should make that quantity available again.

What it enables

Requests move through pending, approved, rejected, and returned states. Approval reduces available stock; returns restore it.

Let the request state control stock changes.

Approval and return actions run inside database transactions. Role-based dashboards separate requesting items from reviewing and managing them.

How I put it together

  • Created a public catalog with availability and borrowing requests.
  • Implemented approval, rejection, return handling, and status notifications.
  • Connected user roles and Google authentication to the application flow.

Tools behind it

Laravel, PHP, MySQL, Blade, Bootstrap, Socialite

Real-time / React

Watch Party

Connected a React video player to room state, host/viewer roles, and real-time playback events through Socket.io.

View source on GitHub (opens in a new tab)

The problem

Separate browsers need a shared playback state. The application also needs to distinguish the person controlling a room from the people joining it.

What it enables

A host can start a room and select its video. Viewers join that room and receive the host timeline, with synchronization metrics available for inspection.

Make the host role explicit.

Starting and joining a room are different actions. The interface follows room state from the server instead of treating every connected browser as a playback controller.

How I put it together

  • Built separate start-room and join-room flows.
  • Integrated MPEG-DASH playback with host-controlled Socket.io events.
  • Added room-state and metrics endpoints, including JSON and CSV exports.

Tools behind it

React, TypeScript, Express, Socket.io, dash.js

Local-first / JavaScript

Local To-Do List

Combined task planning and logbook records with local persistence, media attachments, backup/restore, and Excel export.

View source on GitHub (opens in a new tab)

The problem

Daily tasks and work notes need to stay easy to revisit. A local application also needs an explicit way to move data between browser profiles or devices.

What it enables

Users can organize tasks, attach media, maintain logbook entries, and export their records as JSON backups or an Excel workbook.

Keep text and media in the storage suited to them.

Task and logbook data use localStorage; media attachments use IndexedDB. Explicit backup and restore actions make the boundary of browser-local storage clear.

How I put it together

  • Built task scheduling, priorities, tags, checklists, and completion states.
  • Added a customizable logbook with JSON import/export and native XLSX output.
  • Stored media in IndexedDB and connected a Windows helper for local reminders.

Tools behind it

JavaScript, HTML, CSS, localStorage, IndexedDB, Python