Samuel Girma Megra

Key Skills

Software

Label Studio

Internal/Proprietary Tooling

Top Subject Matter

No subject matter listed

Top Data Types

Computer Code Programming

Image

Text

Video

Top Label Types

Action Recognition

Classification

Computer Programming Coding

Data Collection

Entity Ner Classification

Fine Tuning

Prompt Response Writing SFT

Question Answering

RLHF

Freelancer Overview

I am a software engineer with hands-on experience in creating and curating high-quality datasets for AI model training, particularly in NLP and supervised fine-tuning projects. My work includes generating, annotating, and refining datasets using Python and Pandas, as well as evaluating and improving model reasoning through techniques like RLHF and Chain-of-Thought rationale correction. I have contributed to projects such as hate speech detection for Afan Oromo, where I handled data preprocessing, feature engineering, and model tuning. I am skilled in building data pipelines, analyzing model behaviors, and optimizing AI workflows across various environments. My technical toolkit includes Python, SQL, Docker, and tools for both backend and data-centric development, enabling me to deliver robust solutions for AI-driven applications.

Labeling Experience

Python AI trainer

Label StudioTextEntity Ner ClassificationQuestion Answering

Scope of the Project: Advancing the Gemini model’s reasoning, agentic tool-use, and multi-modal capabilities for complex data processing and workflow automation. Specific Data Labeling Tasks Performed: Engineered SFT datasets (Python/Pandas) for data-to-plot generation; conducted RLHF to correct Chain-of-Thought reasoning; designed multi-step execution paths for simulated workflows (Slack/Jira); and labeled multi-modal data (video/image annotation). Project Size: Managed high-volume, cross-domain datasets spanning code generation, enterprise automation, and visual reasoning. (Tip: Add a specific number here if you have one, e.g., "Curated 5,000+ data points.") Quality Measures Adhered To: Enforced strict RLHF protocols to eliminate hallucinations and utilized "gold-standard" execution walkthroughs to validate high-fidelity tool-call accuracy.

Scope of the Project: Advancing the Gemini model’s reasoning, agentic tool-use, and multi-modal capabilities for complex data processing and workflow automation. Specific Data Labeling Tasks Performed: Engineered SFT datasets (Python/Pandas) for data-to-plot generation; conducted RLHF to correct Chain-of-Thought reasoning; designed multi-step execution paths for simulated workflows (Slack/Jira); and labeled multi-modal data (video/image annotation). Project Size: Managed high-volume, cross-domain datasets spanning code generation, enterprise automation, and visual reasoning. (Tip: Add a specific number here if you have one, e.g., "Curated 5,000+ data points.") Quality Measures Adhered To: Enforced strict RLHF protocols to eliminate hallucinations and utilized "gold-standard" execution walkthroughs to validate high-fidelity tool-call accuracy.

2023

Education

Work History