Abdelghani B.

Key Skills

Software

Don't disclose

Top Subject Matter

Software engineering fundamentals and automated evaluation of programming outputs

AI/LLM RAG pipeline for programming-language documentation

Top Data Types

3D Sensor

Top Task Types

Text Generation

Freelancer Overview

As a systems-focused Software Engineer trained through the rigorous, project-driven environment of 1337 (42 Network), I bring algorithmic precision, structural optimization, and strict technical rigor to AI training data workflows. Beyond manual curation, my experience engineering custom low-level components and context-specific Retrieval-Augmented Generation (RAG) pipelines gives me firsthand insight into how precisely structured data impacts LLM evaluation and semantic alignment. Proficient in C/C++, TypeScript, Rust, and containerized infrastructure (Docker/Proxmox), I am uniquely equipped to autonomously label high-quality technical codebases, programmatically audit data pipelines, and manage data serialization tasks with meticulous attention to detail.

Labeling Experience

Verse Language RAG System — Independent Project

Text Generation

Built an independent Verse language Retrieval-Augmented Generation (RAG) tool designed to ingest, parse, and query context-specific documentation. The work involved configuring semantic search and vector context mappings so that retrieved passages could be used for downstream generation. This is effectively an AI training/pipeline setup where documentation was structured into retrievable context for LLM use. • Ingested and parsed documentation for the Verse programming language into a searchable knowledge base • Implemented semantic retrieval using embeddings with LlamaIndex and ChromaDB • Tuned vector contextual mappings to improve query relevance • Enabled accurate context querying to support RAG-based responses.

2025 - Present

1337 Coding School (42 Network) — Intensive Software Engineering Curriculum (Peer-to-peer learning with strict automated grading)

Don't disclose

Participated in the 1337 (42 Network) peer-to-peer intensive curriculum with project-based automated grading that effectively trains AI-like correctness under strict coding constraints. Activities emphasized systematic debugging and rigorous adherence to standards, which are analogous to generating and validating labeled outputs for programming tasks. The training focused on passing predefined evaluation suites rather than manual review or human annotation workflows. • Completed C and C++ curriculum projects with strict automated checks for memory correctness • Practiced diagnosing bugs and segmentation faults using systematic troubleshooting • Followed coding rules that functioned as quality labels for each submission • Iterated until outputs matched the rubric used by grading automation.

2025 - Present

Education

Work History