Ahmet Hamza Emra

A detection system for CO2 anomalies in the ISS. I co-developed this product in the 2018 Wearable Technologies Workshop under the auspices of NASA and the North American University. I focused on analyzing the causes of CO2 clouds and detecting them from IoT devices. We also created a user interface to gather more data for further research​ on the subject.

A system that uses supervised learning algorithms to classify resumes of top software engineers in various companies and label their skills accordingly. When a user uploads their resume, it is tested against the predicted hypothesis by the ML algorithms, and the system returns what tier(s) the user belongs to.

Open-source lock software with facial recognition and liveness detection. It uses a landmark model for face recognition and a 3D convolution network to predict the liveness of the camera input. It is spoofing-safe and open-source to ensure​ the ​safety of the user.

This project aims to automate the logo creation process using generative AI models and web-scraped data. The proposed approach is based on diffusion models and attention blocks implemented in PyTorch. The system takes user-defined parameters and generates diverse and high-quality logos that meet the constraints. The proposed logo generation system offers advantages over traditional logo design, including speed, cost-effectiveness, and the ability to generate a large number of logo variations. The project demonstrates the potential of AI in automating creative tasks and providing value to businesses and individuals.

In this project, we leveraged the capabilities of a large language model (LLM) to create a personalized AI that simulates the conversational style of a specific individual. To achieve this, we meticulously curated a dataset of our chat history with the individual in question. This dataset served as the training ground for the LLM, enabling it to grasp and reproduce the unique linguistic nuances and patterns characteristic of our friend's communication style. The finetuning process was facilitated by Lora, a cutting-edge technique known for its efficiency in adapting LLMs to specific tasks or styles with minimal data and computational resources. The result is an AI companion that mirrors our friend's conversational idiosyncrasies, offering a new dimension in digital communication that is both personalized and engaging. Although It was successful in mimicking, it also revealed too much personal content, so this project was terminated.