WOLETEMARYAM LIYEW ABITEW

I have been working as AI researcher at GIST for more than 1 year. Interested in building neuro-symbolic model that can leverage the benefits of both the deep learning models and the symbolic one which can balance and get the most out of this models. So was working on developing program synthesizer using neural networks as feature extractor and genetic programming as solution generator. This method leverages domain specific language called DSLS. the program generated by the genetic programming is all symbolic and explainable. At the same time the genetic programming was powered by the neural network specifically vision transformer (DINO) for extracting important features that are helpful for making these solution for a given problem. In the lab using the abstraction and reasoning corpus as a benchmark for measuring the performance of the model.

ML and DL:  (TensorFlow,JAX, Pytorch ,ViT)

Languages used: Python  (Django, Flask, Matplotlib, Seaborn, Pandas), Java, PHP), C++, JavaScript (NodeJS, ReactJS, ExpressJS), C#(.NetCore)

Front-end: HTML, CSS, Bootstrap

Database Management: MySQL, MongoDB

Abductive symbolic solver on the abstraction and reasoning corpus

IJCAI workshop on logical foundations of neuro-symbolic AI, 2024

Korean food detection system: built a website tool in which users can upload images of their model and then it can tell back the name of the food the image has.  
built using Resent for image detection and flask for the backend Integration.
Trip Recommendation System :  designed for travelers in korea that recommends places to visit 

backend built on using flask integrated with custom recommender model

Reinforcement Learning Approach For Solving the 1D-ARC: using reinforcement learning approach for solving the abstraction and reasoning corpus

• Lotte Foundation-GIST Global Scholarship, 09/01/23
• Arcathon 2023, Ranked top 7, 2023

• Neuro-symbolic AI, Program Synthesis ,Evolutionary AI, Genetic Programming