·4 min read
University Projects: From AI to Mobile, Bluetooth to C#
The projects I built during my computer engineering degree, on one page: handwritten digit recognition with a CNN, EduNotice with Flutter and Firebase, PILLDATE with FlutterFlow, an HC-06 Bluetooth remote in React Native, and a Smart Home Manager in C#.
Mission specs
- Handwritten digit recognition
- Python · TensorFlow/Keras CNN · MNIST · 97.60% test accuracy
- EduNotice
- Flutter + Firebase · announcements, files and images from teachers to a class or person
- PILLDATE
- FlutterFlow (no-code) · medication and dose logging, reminders, calendar, dark mode
- MikroControl
- React Native · Bluetooth remote for Arduino over HC-06 (forward/back/left/right/stop)
- Smart Home Manager
- C# / .NET 8 console · rooms, devices and temperature, music via NAudio
- C# assignments
- Assignments 1–4: basic algorithms, classes, arrays and overloading
Team
- Me
Samet GüçlüAll projects
Yusuf Enes ÖztürkSmart Home Manager
Mission log
Every change to the project is logged here with its date, newest first.
- Site
University projects gathered on a single page.
- MikroControl
HC-06 Bluetooth remote app on GitHub: React Native, listing paired devices, connecting and single-letter drive commands.
- EduNotice
EduNotice on GitHub: a teacher–student announcement app built with Flutter and Firebase.
- PILLDATE
PILLDATE announced on LinkedIn: a no-code medication tracker built with FlutterFlow.
Image records
Handwritten digit recognition
Smart Home Manager
Technical report
Architecture, hardware and software in detail.
Introduction
This page gathers the smaller projects I built during my computer engineering degree, for courses and out of my own curiosity. Each one let me try a different area: AI, mobile apps, no-code, controlling hardware over Bluetooth, and object-oriented programming. All the code is on GitHub; the links are at the top of the page.
🖋️ Handwritten digit recognition
A convolutional neural network (CNN) that recognises handwritten digits 0–9. I built it in Python in a Jupyter notebook with TensorFlow/Keras; the goal was to go through data preprocessing, model training and evaluation end to end.
- Data: MNIST, 28×28 greyscale images: 60,000 for training, 10,000 for testing.
- Preprocessing: pixels scaled to 0–1 and images reshaped to
28×28×1for the CNN. - Model: two convolution + pooling layers (32 and 64 filters), then a 128-neuron dense layer and a 10-class softmax output.
- Training: Adam optimiser,
sparse_categorical_crossentropyloss, 5 epochs. - Result: 97.60% accuracy on the test set. The trained model is saved as
model.h5.
| Epoch | Training accuracy | Test accuracy |
|---|---|---|
| 2 | 93.97% | 95.27% |
| 3 | 95.79% | 96.45% |
| 4 | 96.56% | 96.88% |
| 5 | 97.03% | 97.60% |
📢 EduNotice
A mobile app that lets teachers send announcements to students quickly and in an organised way. I wrote it in Flutter and used Firebase for authentication, the database and file storage.
- Roles: teachers and students sign in separately. Teachers post announcements; students pick their class when they register.
- Targeting: an announcement can go to everyone, to specific classes or to selected students, and each student only sees what's meant for them.
- Content: besides a title and text, announcements can carry an image and multiple files, uploaded to Firebase Storage.
- Tracking: unread announcements are marked and the list is sorted by date.
- UI: Material Design, with light and dark modes that follow the system theme.
Tech: Flutter · Firebase Authentication · Cloud Firestore · Firebase Storage · file_picker / image_picker.
💊 PILLDATE
A medication tracker I built with a no-code approach in FlutterFlow. My aim was to use the speed of no-code tools to solve an everyday problem: helping people manage their treatment simply and reliably, without missing doses.
- Quick entry: medication name, dose amount and unit are recorded in seconds.
- Scheduled reminders: notifications at set times reduce the risk of forgetting a dose.
- Daily and weekly view: a calendar screen shows which medication to take on which day; doses can be scheduled every day, every few days or on specific weekdays.
- Dark mode: a modern, comfortable interface.
🎮 MikroControl: an HC-06 Bluetooth remote
An Android app I wrote in React Native to drive an Arduino-based vehicle from a phone. It connects to an HC-06 Bluetooth module over classic Bluetooth (SPP).
- Checks that Bluetooth is on and that the required permissions are granted.
- Lists devices paired with the phone, and can connect to and disconnect from the selected one.
- The direction buttons send single-letter commands:
Fforward,Bback,Lleft,Rright,Sstop. - The last command sent and the connection state are shown on screen, and errors are reported clearly to the user.
Tech: React Native 0.84 · react-native-bluetooth-classic · Android.
🏠 Smart Home Manager
A C# console application that simulates home automation, which I built with Yusuf Enes Öztürk. The starting point: a home's electronic devices and room temperatures couldn't be controlled from a single place.
- User sign-in: a sign-up and sign-in flow.
- Rooms: add and remove rooms; each room's temperature is changed with validation to the 15–30 °C range.
- Devices: add, remove, switch on and off, and list the state of the devices in each room.
- Sound system: play and stop the MP3s in a music folder using the NAudio library.
- Code structure: a modular, reusable design with
Oda(room),ElektronikEsya(electronic device) andSicaklikKontrol(temperature control) classes;Odainherits fromSicaklikKontrol.
Tech: C# · .NET 8 · Visual Studio · NAudio.
📚 C# assignments
My first assignments in the introductory programming course (October–November 2024): addition, multiplication, division and modulo, the volume of a prism, odd/even checks, converting numbers to words with switch, lists and sums, counting words in a sentence, sorting, the area of a triangle, the largest value in an array, and method overloading. Small exercises, but the foundation for every project that came after.
🎯 Conclusion
Each of these projects taught me something different: the importance of data and evaluation in AI, role and permission separation in mobile apps, shipping quickly with no-code tools, sending commands from software to hardware over Bluetooth, and the fundamentals of object-oriented design. Most of my later, larger projects were built on top of these small experiments.