Maybemar Hernandez

Key Skills

Software

Clickworker

Remotasks

Scale AI

Toloka

Top Subject Matter

No subject matter listed

Top Data Types

Computer Code Programming

Image

Text

Top Task Types

Computer Programming Coding

Evaluation Rating

Prompt Response Writing SFT

RLHF

Freelancer Overview

I am a software engineer, English and Spanish speaker, that have worked in more than 50 projects for AI Training. Design advanced prompts to evaluate the functionality and limitations of LLMs to process human language and understand coding requirements. Design prompts and develop the perfect response in Python for SFT projects. Review and fix code in Python, and other programming languages (Js, SQL, C++, etc..) to provide feedback, in RLHF projects. Design advance prompts and perfect responses in SFT projects to beat SOTA models in UI development in HTML/CSS/JS and Tailwind CSS. Analyze and evaluate prompts and perfect responses in Spanish. Drafting and revision of rubrics in English and Spanish for model fine tuning. Evaluation of question and answers for risk evaluation for safety in model responses. Squad reviewer. Leaded teams of contributors to ensure quality of their work in the ongoing projects.

Labeling Experience

Translations

Scale AITextText Generation

Evaluate and fix the prompt and response translations in Spanish keeping the meaning of both.

Evaluate and fix the prompt and response translations in Spanish keeping the meaning of both.

2024

Safety Evaluation

Scale AITextEvaluation Rating

Evaluate prompt and responses in Spanish to determine if the model provided the specific punts to unsafe requests including jailbreak requests.

Evaluate prompt and responses in Spanish to determine if the model provided the specific punts to unsafe requests including jailbreak requests.

2024

UI Project

Scale AIComputer Code ProgrammingPrompt Response Writing SFT

Create advanced prompts with high-level constraints to request UIs for web and mobile apps built in HTML/CSS/JS and Tailwind. Create a perfect response in these languages and frameworks beating the SOTA models, and providing a 500 fortune level UI.

Create advanced prompts with high-level constraints to request UIs for web and mobile apps built in HTML/CSS/JS and Tailwind. Create a perfect response in these languages and frameworks beating the SOTA models, and providing a 500 fortune level UI.

2024

Code Extensions

Scale AIComputer Code ProgrammingRLHF

Using RLHF between two model responses to determine if the usage of external extensions APIs (for example Google APIs) was correctly used to respond to specific user prompts.

Using RLHF between two model responses to determine if the usage of external extensions APIs (for example Google APIs) was correctly used to respond to specific user prompts.

2024 - 2024

Spanish Rubrics

Scale AITextEvaluation Rating

Create a prompt in Spanish and evaluate the model responses creating rubrics with what an ideal response should have considering all the requirements of the prompt. Evaluation included instruction following, accuracy, language, and cultural knowledge.

Create a prompt in Spanish and evaluate the model responses creating rubrics with what an ideal response should have considering all the requirements of the prompt. Evaluation included instruction following, accuracy, language, and cultural knowledge.

2024 - 2024

Education

Work History