Milos Lupurovic

Logger is an ESP8266-based device running on FreeRTOS with firmware developed in C++. It is designed to monitor and control solar power plants by communicating with inverters, meters, and sensors, while providing real-time data to a web platform and supporting communication with grid operators. The device uses Modbus (RTU/TCP), MQTT, and WebSockets protocols to ensure reliable data acquisition, remote control, and seamless integration into monitoring and control systems.

I developed a medical device powered by an ARM-based processor running a custom Linux distribution built with the Yocto Project. I designed and implemented a C++ provisioning service responsible for securely onboarding devices to AWS IoT. The service communicated with a mobile application via BLE and NFC to perform device provisioning and identity exchange. Additionally, it managed the AWS IoT Device Shadow, enabling users to remotely read, update, and control device configuration and operational behavior through the cloud.

The Validator is a hardware component of an automated fare collection (AFC) system for automating the ticketing system of a public transportation network. It's a system-on-a-chip (SoC) device based on a custom Linux operating system with additional peripherals, including a contactless smart card reader. The Validator is a level 1 device of the AFC system and complies with the SDOA standard.

Additionally, the Validator can communicate with the devices on the same vehicle via various protocols such as TCP, UDP, HTTP, and the Azure Cloud for sending status and receiving commands.

Integration of ESP8266 chip into heat pump to enable remote monitoring and control of the pump via Azure IoT Hub.
ESP8266 was connecting to the Azure Cloud from where it received various commands, which were sent to the pump motherboard using RS-232 protocol. ESP8266 chip was also reading the status of the pump and was periodically sending that status to the Azure Cloud.

A recommender system featuring keyword extraction, summary generation, vectorization and classification, article search with a custom re-ranking and suggesting mechanism, profile scoring, user activity tracking, and article recommendations based on desired categories and user activity. I built the pipeline that analyzes, processes, and cleans the articles as a member of the team that developed the system.

Implemented Pomodoro Timer in the form of a desktop application. Pomodoro is a time management technique where work is focused on 25-minute periods, followed by a short or long break, depending on the number of previously finished cycles. This desktop app was implemented using C++ and Qt. It was a cross-platform solution supported on Windows, macOS, and Linux. Additionally, the application was able to communicate with a custom-made USB HID device, which was a small LED light, to set the color of the light, indicating the current Pomodoro status.

Integrated a Raspberry Pi board into a small four-wheeled rover controlled via four DC motors with integrated rotatory encoders and the purpose of the Raspberry Pi was to listen for user input and transform it into commands used to control the motors while utilizing the feedback from the encoders. The drive was designed to prevent wheel slippage and minimize positioning errors. Communication between users and Raspberry Pi could be done via local MQTT communication or AWS cloud.
• Rover could be controlled by voice commands using Amazon Alexa.
• Wall detection using a proximity sensor enabled the rover to stop if it stumbled upon obstructions in its path.
• A front-facing camera on the rover enabled 1st person view by having a video streaming server on the Raspberry Pi itself