Projects

This project explores the effectiveness of YOLO-based systems for drone detection across multiple datasets using various data augmentation strategies. The main focus is detecting drones under different conditions—from distant, cloudy shots to closer, higher-resolution frames. By training and evaluating four separate models with different augmentation techniques (horizontal/vertical flipping) and combining two distinct datasets (Anti-UAV and VisioDECT), we analyzed how these factors impact detection performance. The study reveals insights into the benefits and limitations of each approach and data domain, achieving up to 99% precision and recall with the best model.

A modular IoT-based smart agricultural system designed to optimize resource management and sustainability through automation. This proof-of-concept demonstrates distributed sensor nodes communicating via LoRa, an ESP32 gateway with secure SSL/TLS connectivity, and a Java server integrated with Kafka for real-time data processing.

This project describes the design and implementation of a robot that uses the STM32F303 Discovery board to control hardware and a Python server application with OpenAI GPT API to process audio signals and generate commands. The main goal is to create an autonomous system that can process audio commands using artificial intelligence and convert them into appropriate signals for motor control. This system enables integration of embedded systems with modern natural language processing, using a server application for communication between user commands and the robot's motor control.

Recreated AMD's 1975 Am2901 4-bit ALU slice in VHDL using Vivado, simulating basic arithmetic and logic operations to explore how modular bit-slice processors worked.

End-to-end telemetry system with LoRa nodes (SX127x), ESP32 gateway, MQTT/Kafka backend, AES-128 encryption, and custom TDMA scheduling for multi-month field operation.

CAN bus analysis on live vehicles (Fiat, Renault) using STM32MP1 and SocketCAN, including UDS reverse-engineering, physical layer characterization, and multi-make diagnostics platform development.

Small outdoor robot for lawn-scale navigation using ultrasonic sensors and a perimeter-wire guidance system with 7kHz inductive detection for reliable boundary tracking.

STM32F303-based differential-drive robot with bare-metal C firmware, DMA-backed UART, center-aligned PWM motor control, and OpenAI Whisper voice command integration.

DIY filament extruder using RAMPS 1.4 board, dual 3D printer nozzles, and NEMA 17 stepper to convert shredded PET bottles into usable 3D printer filament.

Remote-trigger humane trap designed for lynx conservation, successfully deployed in Mavrovo National Park for wildlife monitoring and research.