Milena Trajanoska

Student, Machine Learning & Data Science enthusiast


Passionate about applied ML & DS research in real-life scenarios
Interested in healthcare and life sciences
Optimizing business processes through data-driven decisions
See my work
Work Experience
Machine Learning & Back-End Engineer at Loka, Inc., Skopje, Macedonia
November 2021 – Present

Helping start-ups form Silicon Valley develop high-tech products and empowering technological advancements. Developing Machine Learning and Deep Learning - based solutions with great focus on NLP including transformer architectures, deep neural networks and statistical methods.  Implementing full-scale infrastructures for Machine Learning applications using AWS services, extensive focus on Amazon SageMaker. Implementing MLOps pipelines for model parameter-tuning, model training and validation with metrics and artifacts versioning.

Technology stack: AWS, Python, TensorFlow / Keras, PyTorch, Sklearn, SQL

Part-time Student Researcher - Data Science at Macedonian Academy of Sciences and Arts, Skopje, Macedonia
August 2021 – Present

Part time student researcher in the field of  Data Science, Machine Learning, and Deep Learning. Assessing macroeconomic issues of European countries, with special emphasis on the Republic of Macedonia. Contributing towards the improvement of institutional structures in my country of origin through implementing state-of-the-art analysis and modeling. The results from the analysis are used by the Economic Chamber of North Macedonia in their decision making process.


Technology stack: Python, TensorFlow / Keras, Sklearn, LaTeX

Education
Faculty of Computer Science and Engineering, Ss. Cyril and Methodius University, Skopje, Macedonia
October 2018 – Present

Hard-working undergraduate student, in my final year of studies. Very enthusiastic about the latest technological developments in the fields of AI and ML.  Current top student of the entire generation with a GPA of 9.88.


Expected graduation date: October 2022.

Internships
Machine Learning Intern at Loka Inc., Skopje, Macedonia             
July 2021 – October 2021

Implementing high-performing image classification Deep Learning models. Utilizing transfer learning for state-of-the-art DL models such as MobileNetV2 and VGG16. Leveraging neural network ensemble methods including: majority voting, Bayesian weighted averaging and integrated model stacking. Developing mobile application using Google's Flutter Framework and developing back end API using Python with the Flask Framework. Creating complete model pipeline for incremental re-training and deployment using Amazon SageMaker.

Technology stack:  Python, Keras, Tensorflow, Flutter, Flask, Amazon SageMaker.

Software Engineer Intern at Netcetera DOOEL, Skopje, North Macedonia
August 2020 – October 2020

Implementing new features for a micro-services platform aiming to empower actions towards sustainable environmental development, Pulse Eco. Unit testing the services layer and manual testing of the user interface. Bug fixing and security improvements. Implementing smart internationalization on two languages: English and Macedonian.

Technology stack: Java Spring Boot, Maven, JUnit, Mockito, Git, JavaScript, Docker, JQuery, CSS, D3.js

Scientific Articles and Projects


Dietary, comorbidity, and geo-economic data fusion for explainable COVID-19 mortality prediction

Soon to be published in the journal Expert Systems with Applications. The Impact Factor of this journal is 6.954, ranking it 24 out of 273 in Engineering, Electrical & Electronic.

More details about scientific articles upon request.




FedCSIS 2022 Challenge: Predicting the Costs of Forwarding Contracts

Ranked 10th out of 135 teams from 24 countries around the world.




Basic Movie Recommendation System

Comparison of the performance of vanilla matrix factorization for collaborative filtering and content-boosted collaborative filtering, including embeddings from movie plots generated using TF-IDF.

Recommendation System for the MovieLens 100K dataset based on matrix factorization techniques for collaborative filtering.

Automobile and real-estate price estimation

Web-crawler architecture
Results for the housing price assessment



Semantic Image Segmentation

Basic image segmentation of the Oxford pets benchmark dataset.

Using convolutional neural networks with a U-net architecture for producing segmentation maps.



Example result: