Amanuel M.

Key Skills

Software

No software listed

Top Subject Matter

AI/LLM infrastructure

Backend engineering

Distributed systems

Top Data Types

Computer Code Programming

Image

Text

Top Task Types

Evaluation/Rating

Computer Programming/Coding

Prompt + Response Writing (SFT)

RLHF

Fine-tuning

Freelancer Overview

My AI training and data labeling experience is centered around my work in LLM Evaluation and Alignment with Revelo. In this role, I evaluated over 50 AI-generated pull requests across real-world Python repositories, scoring them on correctness, test coverage, and edge-case handling. I performed structured code audits by comparing multiple model outputs to produce high-quality labeled feedback. This feedback was directly utilized in LLM alignment workflows to improve coding models. Complementing my direct data labeling experience is my technical background in engineering AI infrastructure. I built DocuMind, an AI Document-Intelligence Platform, where I architected a multi-tenant RAG system that grounds LLM answers in user documents. Through engineering asynchronous ingestion pipelines and managing vector embeddings, I possess a holistic understanding of how labeled training data and backend architectures integrate to create robust AI systems.

Labeling Experience

Anthropic Post-Training Pipeline: RLHF & Code Alignment

Computer Code ProgrammingRLHF

In this role, I contributed to the post-training alignment pipelines for Anthropic’s LLM models via Revelo. My core responsibility was to evaluate and perform structured code audits on 50+ AI-generated pull requests and complex programming tasks. I scored outputs based on logical correctness, test coverage, and edge-case robustness. By comparing multiple model outputs, I produced highly curated, labeled feedback that was directly integrated into RLHF and SFT workflows to improve model performance and reliability in coding tasks. This work required deep technical analysis of Python codebases to ensure the model's generated logic adhered to industry best practices and security standards.

2024 - Present

Education

Work History