Fidel Caicedo

Key Skills

Software

Data Annotation Tech

Scale AI

Top Subject Matter

No subject matter listed

Top Data Types

Computer Code Programming

Image

Text

Top Task Types

Text Generation

RLHF

Fine Tuning

Freelancer Overview

I am an experienced backend engineer with a strong background in AI training data and reinforcement learning from human feedback (RLHF) workflows. My work has centered on developing and reviewing backend code examples in Python and TypeScript to evaluate and enhance large language model (LLM) code quality. I am skilled in data annotation, code evaluation, and automation, with hands-on experience integrating and maintaining scalable backend services using GraphQL, Node.js, and FastAPI. My focus on secure, reliable releases through CI/CD pipelines, automated testing, and DevSecOps practices ensures high-quality data and robust systems. I am passionate about improving AI systems by ensuring data integrity, consistency, and quality throughout the training process.

Labeling Experience

AI Training Specialist (RLHF Model Optimization)

Scale AIComputer Code ProgrammingText GenerationRLHF

As an AI Training Specialist at Outlier, I produced and reviewed backend-focused code samples and tests to train and evaluate LLM coding behavior via RLHF. I improved generated code correctness by applying engineering best practices, including input validation, error handling, and edge-case coverage. Rubric-based reviews were used to ensure high-quality, diverse code generation for model improvement. • Generated and curated code samples in Python and TypeScript • Applied comprehensive review standards for LLM evaluation • Ensured code reliability and accuracy for RLHF workflows • Focused on improving the quality of code generation in LLMs

As an AI Training Specialist at Outlier, I produced and reviewed backend-focused code samples and tests to train and evaluate LLM coding behavior via RLHF. I improved generated code correctness by applying engineering best practices, including input validation, error handling, and edge-case coverage. Rubric-based reviews were used to ensure high-quality, diverse code generation for model improvement. • Generated and curated code samples in Python and TypeScript • Applied comprehensive review standards for LLM evaluation • Ensured code reliability and accuracy for RLHF workflows • Focused on improving the quality of code generation in LLMs

2025

Education

Work History