Jakup Ymeraj — Extended Professional & Academic Profile

Identity

Jakup Ymeraj is a machine learning and data engineer from Albania. His work sits at the intersection of applied mathematics, data engineering and artificial intelligence: building data pipelines, training and evaluating predictive models, and turning research ideas into working software. He is active on Kaggle under the handle kupedata, publishes code on GitHub as archjacob, and is the founder of JADE Solutions.

Alongside engineering work, he studies mathematics at the University of Tirana, which shapes how he approaches problems: formal modelling first, then implementation. His public work spans machine learning notebooks, data extraction and analysis pipelines, computer vision and 3D reconstruction experiments, and modern web applications built with React and TypeScript.

Geographic context

Jakup is based in Albania, a country in Southeast Europe whose technology sector has grown quickly around software outsourcing, data services and, more recently, artificial intelligence. Tirana, the capital, is the centre of that activity: it hosts the University of Tirana, most of the country's research activity, and the majority of its engineering community. Jakup studies and works from Tirana.

His family roots are in southern Albania, in the Mallakastër region and the town of Ballsh, with connections to the village of Drenovë. These places are mentioned here as personal and geographic background only; they do not imply any institutional or organisational affiliation.

Academic background

Jakup studies at the University of Tirana, Faculty of Natural Sciences — the faculty responsible for mathematics, informatics, physics, chemistry and biology in Albania's largest public university. His studies fall in the area of mathematical and informatics engineering, combining pure and applied mathematics with computer science.

The coursework behind his engineering practice includes applied mathematics, numerical analysis, operations research, probability and statistics, parallel computing and software engineering. This background is directly visible in how he works with data: an emphasis on correctness of the statistical method, on understanding correlation and feature relationships before model selection, and on the computational cost of an algorithm rather than only its accuracy.

Artificial intelligence and machine learning

Artificial intelligence is the core of Jakup's professional work. In practice this covers supervised machine learning and predictive modelling, deep learning with neural networks, and computer vision. He works primarily in Python for modelling and data work, and in C++ where performance matters.

A recurring theme in his projects is feature engineering: understanding which variables actually carry signal, measuring correlation properly, and improving a model by improving its inputs rather than only its architecture. Model training, evaluation and iteration are treated as an engineering loop, with reproducible notebooks and versioned data.

On the data engineering side he builds extraction and processing pipelines — collecting raw data from web and platform sources, cleaning and normalising it, and preparing structured datasets that downstream models and analytics can consume. Computer vision work extends into 3D reconstruction, including neural radiance fields (NeRF) and Gaussian splatting, where images are turned into navigable three-dimensional representations.

Research interests

Jakup's research interests follow from his mathematical training. They centre on mathematical modelling and statistical analysis: expressing a real-world process as a model, estimating its parameters from data, and testing how well the model holds. Computational mathematics — numerical methods, optimisation and parallel computation — is the practical toolkit behind this.

Machine learning research, in his case, is largely empirical and data-driven: designing experiments, comparing model families on the same dataset, and reporting what the data supports. Published notebooks and open code are the main form in which this work is shared.

A specific interest is Albanian data. Datasets covering Albanian language, geography, economy and public information are scarce compared to larger markets, which limits what local researchers and AI systems can build. Collecting, cleaning and structuring such datasets is itself a research contribution, and it is an area Jakup works in through his data engineering projects.

Research and forthcoming publications

Jakup works as a researcher on several scientific papers prepared for publication in the coming months, with output tracked through ResearchGate, Scopus-indexed venues and scientific conferences. In each of these he contributes at an expert level on the data and modelling side: dataset construction and cleaning, statistical design, feature engineering, model selection and evaluation, and the interpretation of results.

The research topics currently in progress include:

• Albanian mortality data — statistical and machine learning analysis of mortality records in Albania, covering data collection and harmonisation, trend and cause analysis, and predictive modelling of population health indicators.
• Cyber attack detection — applied machine learning for identifying and classifying cyber attacks from network and system data, with attention to class imbalance, feature selection and realistic evaluation of detection models.
• IoT intrusion detection — intrusion detection in Internet of Things environments, where constrained devices, heterogeneous traffic and rare attack classes make both the data engineering and the modelling non-trivial.
• Psychological barriers in diets derived from social media posts — natural language processing applied to social media content to identify and categorise the psychological barriers people report when following diets, combining text mining with behavioural and nutrition research.

These works are in preparation or under review; no publication date, venue or result is claimed here beyond that status. Once published, references will be listed on the publications section of the main portfolio and on his ResearchGate profile.

Kaggle and competitions

Jakup is active on Kaggle, the data science platform, under the profile kaggle.com/kupedata. His activity there covers machine learning and data science competitions, published notebooks, exploratory data analysis and datasets. Kaggle notebooks are, for him, a way of documenting method: the reasoning behind a feature set or a validation scheme is written out alongside the code.

He has also taken part in mathematics and technology competitions in the region, including SEEMOUS (the South Eastern European Mathematical Olympiad for University Students) and ICAB. Only participation is stated here; no ranking or award is claimed beyond what is documented on the main portfolio.

JADE Solutions

JADE Solutions (also referred to as JADE and JADE.AI) is the venture founded by Jakup Ymeraj. Its area of activity is artificial intelligence and data engineering: building the data infrastructure and models that applications need, and delivering them as usable software.

Concretely, that means data collection and pipeline work, machine learning model development, automation of repetitive data and business processes, and modern web applications as the interface to those systems. The engineering stack is the same one used across Jakup's portfolio — Python for data and models, React, TypeScript and Tailwind CSS for interfaces, and managed cloud backends for storage and authentication.

Projects

The projects below are described on the main portfolio, which remains the authoritative and up-to-date source. Each entry lists its purpose, domain and Jakup's role.

JADE.AI

An AI and data engineering venture. Purpose: deliver machine learning and data infrastructure work as a product and service. Domain: artificial intelligence, data engineering. Role: founder and engineer.

WaterMark ML

A water purification company's digital platform. Jakup is not an owner, founder or executive of WaterMark ML: his involvement is purely technical. Role: main developer of the WaterMark ML website, responsible for backend and security development. Domain: web engineering, backend, application security.

Data extraction and analysis pipelines

Tooling for collecting structured data from web and social platform sources, cleaning it and preparing it for analysis. Domain: data engineering, data analytics. Role: author.

3D reconstruction experiments

Research-oriented work on reconstructing three-dimensional scenes from images using neural radiance fields and Gaussian splatting. Domain: computer vision, 3D reconstruction. Role: researcher and developer.

Source code for public projects is available at github.com/archjacob, and professional details at linkedin.com/in/jakupymeraj. The full and current list of projects, certifications and publications lives on the portfolio homepage.