Mardem A.

Key Skills

Software

Other

Top Subject Matter

Open-Set Recognition & Computer Vision

Large Language Models (LLMs) & NLP

Machine Learning & Python Code Evaluation

Top Data Types

Text

Image

Computer Code Programming

Top Task Types

Fine-tuning

Classification

Evaluation/Rating

Bounding Box

Object Detection

Text Summarization

Transcription

Freelancer Overview

Computer Science Master's candidate at Universidade Federal de Minas Gerais (UFMG) specializing in Open-Set Recognition and Computer Vision. Brings academic rigor and professional experience in Machine Learning workflows, including hierarchical classification and data preprocessing. Core strengths include Python, C/C++, PyTorch, and NLP modeling. Proven track record in AI evaluation, with hands-on experience fine-tuning LLMs (DistilBERT) and engineering Transformer architectures from scratch. AI-training focus includes Computer Code, Programming, and Text workflows, with deep expertise in model evaluation, advanced fine-tuning, and complex output rating.

Labeling Experience

Undergraduate Researcher (Paid) - Universidade Federal de Lavras

ImageClassification

You served as an undergraduate researcher working on machine learning and computer vision problems under a paid research arrangement. You researched and proposed feature-selection techniques for hierarchical classification and adapted the Inconsistency Rate metric to hierarchical contexts. You authored key parts of an upcoming academic paper, translating structural methodology into algorithmic results. • Researched feature selection for hierarchical classification • Adapted Inconsistency Rate from flat to hierarchical settings • Produced manuscript content as primary author • Applied computer vision and ML research methods

2024 - 2025

Undergraduate Researcher (Paid) — hierarchical classification feature selection and metric adaptation

OtherTextFine-tuning

Conducted machine learning research focused on improving hierarchical classification performance. Adapted a feature/filter metric (Inconsistency Rate) from a flat learning setup to a hierarchical context to support better decision consistency. Contributed to the development of an academic paper describing the structural methodology and algorithmic results. • Designed or adapted hierarchical classification evaluation using the Inconsistency Rate metric • Authored an upcoming academic paper detailing methodology and results • Investigated feature selection techniques for hierarchical classification problems • Worked in a research setting applying AI methods to structured label spaces

2024 - 2025

LLM Fine-Tuning for NLP Classification — DistilBERT transfer learning

OtherTextFine-tuning

Fine-tuned a pre-trained DistilBERT model for NLP classification tasks using transfer learning. Adapted the model to specialized text datasets for multi-class emotion detection and fake news identification. Applied tokenization and training strategies to improve predictive accuracy. • Fine-tuned DistilBERT for multi-class emotion detection • Fine-tuned for fake news identification • Used transfer learning and tokenization to adapt general models • Optimized accuracy for specialized textual datasets

2024 - 2024

Transformer Architecture Implementation — from-scratch implementation in Python/PyTorch

Other

Implemented Transformer neural network components from scratch to learn and demonstrate core deep learning mechanisms. Built multi-head self-attention and positional encoding directly in code rather than using off-the-shelf implementations. Produced a complete architecture implementation for understanding modern LLM foundations. • Implemented multi-head self-attention and positional encoding • Developed a complete Transformer model implementation from scratch • Validated the architecture understanding through a full codebase • Released/maintained the project via a GitHub repository

2024 - 2024

Undergraduate Researcher (Volunteer) — ML analysis, preprocessing, and predictive evaluation for cultivation data

OtherText

Applied machine learning methods to analyze and preprocess sweet potato cultivation datasets and extract predictive insights. Performed extensive data preprocessing and evaluated predictive accuracy across multiple machine learning models. Authored a peer-reviewed paper related to the cultivated dataset analysis workflow. • Conducted data preprocessing to prepare inputs for ML models • Trained/evaluated multiple machine learning models for predictive accuracy • Extracted key insights from sweet potato cultivation data • Co-authored/primary-authored a peer-reviewed publication (RITA journal)

2023 - 2024

Education

Work History