About Me

Ephrem Mekonnen

Hi, I am Ephrem, a PhD student at Technological University Dublin, where I work on the explainability of deep neural networks. Previously, I worked as a Research Intern at Bruno Kessler Foundation for eight-plus months. Currently, I am also participating in the machine learning part of the HappyMaths project led by my supervisor ( Dr. Pierpaolo Dondio)
My research interests are Explainable AI, Deep Learning, and Machine Learning.

Experiences

LIFETOUCH srl

May 2022 - July 2022
Software Engineer

PerVoice SpA

March 2022 - April 2022
Internship Trainee

Fondazione Bruno Kessler - FBK

February 2021 - October 2021
Research Intern

Haramaya University, Ethiopia

September 2017 - August 2019
Higher education teaching professional

Ethiopian Airlines

July 2016 - September 2016
Software developer Intern

Education

TU Dublin, Ireland

October 2022 - September 2026
PhD Student in Computer Science

University of Trento, Italy

September 2019 - December 2021
Master's Degree in Computer Science

Addis Ababa University, Ethiopia

September 2013 - July 2017
Bachelor of Science Degree in Computer Science

Skills

My Projects

End to End Keyword Spotting (E2E KWS)

An end-to-end low-resource keyword spotting through a characterlevel LSTM and connectionist temporal classification loss function

...   ...

Mining Frequent Popular Topics in time from tweet dataset

Given the tweet dataset about covid-19, and the idea of the project is to find frequent popular words that appear together in time. It is implemented fully in python.

Quorum-Based Total Order Broadcast

Implementing a protocol for the coordination of a group of replicas, sharing the same data, that tolerates multiple node failures thanks to a Quorum-based approach.The system handles update requests coming from external clients and guarantees all replicas apply the update in the same order. Updates are managed by a special replicas, the coordinator, following a two-phase broadcast procedure. if the coordinator crashes, a new one is elected to ensure continuity of service. It is coded in Java (Using Akka framework)

...   ...

Pharmacy Inventory Management System

A desktop application that helps to register stocks, manage sells and purchases easily. It has well-implemented account management so that mangers of the pharmacy can control what is going on and take efficient decisions accordingly. The application can also support different languages.

Deploying Web Application using Docker and Kubernetes

Fog and Cloud Computing Class project. Deploying web application using docker and kubernetes to easily scale up/ down the web server as per the traffic, mange and leverage the benefits of both technologies.

...   ...

Certifications


Publications

ICASSP

END-TO-END LOW RESOURCE KEYWORD SPOTTING THROUGH CHARACTER RECOGNITION AND BEAM-SEARCH RE-SCORING

Mekonnen, Ephrem; Brutti, Alessio and Falavigna, Daniele

In 2022 IEEE International Conference on Acoustics, Speech, & Signal Processing (ICASSP)

XAI-2023

Explaining Deep Learning Time Series Classification Models using a Decision Tree-Based Post-Hoc XAI Method

Mekonnen, Ephrem; Dondio, Pierpaolo and Longo, Luca

xAI-2023 Late-breaking Work, Demos and Doctoral Consortium Joint Proceedings

Frontiers-in-AI

A Global Model-Agnostic Rule-Based XAI Method based on Parameterised Event Primitives for Time Series Classifiers (Under Review)

Mekonnen, Ephrem; Dondio, Pierpaolo and Longo, Luca

Frontiers in Artificial Intelligence Journal

XAI-2024

Interpreting Black-Box Time Series Classifiers using Parameterised Event Primitives

Mekonnen, Ephrem; Longo, Luca and Dondio, Pierpaolo

xAI-2024 Late-breaking Work, Demos and Doctoral Consortium Joint Proceedings

© 2024 Ephrem Mekonnen