Youssef M.

Key Skills

Software

Don't disclose

Top Subject Matter

Video super-resolution (spatio-temporal reconstruction)

LLM post-training (SFT, QLoRA, DPO) and agentic evaluation

Top Data Types

Video

Image

Text

Top Task Types

Fine-tuning

Segmentation

Bounding Box

Object Detection

Question Answering

Text Generation

Computer Programming/Coding

Transcription

RLHF

Freelancer Overview

Research Intern, Knights Lab (AI model evaluation and baseline replication for video super-resolution). Brings 1+ years of professional experience across complex professional workflows, research, and quality-focused execution. Core strengths include Agentic Programming, Generative AI (diffusion models & GANs), programming, software development, LLM finetuning, and Hugging Face. Education includes a Bachelor of Science from Alexandria University (2026). AI-training focus includes data types such as Computer Code and Programming, and labeling workflows including Evaluation, Rating, and Fine-tuning.

Labeling Experience

LLM Post-Training & Agentic Evaluation Pipeline (project-based AI training)

Fine-tuning

Implemented an academic LLM post-training pipeline for supervised fine-tuning and preference optimization. Configured memory-efficient training with low-bit quantization and mixed-precision to support limited hardware. Evaluated post-training behavior changes related to instruction following, code generation, and safety. • Supervised fine-tuning, QLoRA, and Direct Preference Optimization on Qwen2-1.5B-Instruct • Training configuration using 4-bit quantization, gradient checkpointing, paged AdamW, and mixed precision • Behavioral evaluation after fine-tuning and preference optimization • Assessed instruction-following, code-generation behavior, and safety-related responses

2021 - 2026

Research Intern, Knights Lab (AI model evaluation and baseline replication for video super-resolution)

Don't disclose

Conducted AI research to compare supervised learning and state-space modeling approaches for video super-resolution. Focused on assessing baseline replication strategies for spatio-temporal reconstruction tasks using existing paper specifications. Worked on benchmarking-related methodology rather than producing human-annotated labels for a dataset. • Reviewed candidate baseline approaches and compared modeling strategies • Set up replication for an unavailable-source Mamba-based video super-resolution architecture • Evaluated model behavior for video reconstruction quality drivers • Documented experiment setup aligned to research specifications

2025 - 2025

Education

Work History