Emiliano Armango

Key Skills

Software

AWS SageMaker

Remotasks

Scale AI

V7 Labs

Top Subject Matter

No subject matter listed

Top Data Types

Audio

Computer Code Programming

Image

Top Task Types

Computer Programming Coding

Evaluation Rating

RLHF

Text Generation

Text Summarization

Freelancer Overview

Actually, my first approach to IA training was 20 years ago: studying programming, the year project was a "smart TicTacToe" that had to learn from each past game. I developed in C# using only dynamic memory, since the main goal was understanding how it learned. Many years later, around 2017 when streaming was not as common as today, I used TensorFlow and OpenCV to set a real-time human detector as part of the R&D team of a banking security company. Finally, for almost 2 years, I have worked as a freelancer with platforms like SageMaker, Scale, V7, and others. I like to work on STEM, coding, and strategic thinking projects, but I enjoy even more designing and refining a pipeline for tasking, as well as setting QA parameters.

Labeling Experience

Contributor

Aws SagemakerComputer Code ProgrammingComputer Programming Coding

The project involves a public repository and specific issues. The goal is to trace the thoughts a programmer has when: creates TDD cases, fixes an issue, and deploys a project.

The project involves a public repository and specific issues. The goal is to trace the thoughts a programmer has when: creates TDD cases, fixes an issue, and deploys a project.

2025

Contributor

Scale AIAudioText Generation

A model recognized an audio clip and described it. The goal was to evaluate the text generated and improve it.

A model recognized an audio clip and described it. The goal was to evaluate the text generated and improve it.

2025 - 2025

Contributor, Reviewer

Scale AIComputer Code ProgrammingComputer Programming Coding

The project involved public repositories, coding agents, and Docker testing environments. The goal was to guide the model by prompts to create certain files (a Dockerfile, python, and bash files mainly) that were enough to build and test the provided repo.

The project involved public repositories, coding agents, and Docker testing environments. The goal was to guide the model by prompts to create certain files (a Dockerfile, python, and bash files mainly) that were enough to build and test the provided repo.

2025 - 2025

Contributor, QM, Consultant

Scale AIImagePrompt Response Writing SFT

The main project I have worked on in this area required an Image, a prompt, a ground truth answer, and an explanation. The model must have failed to solve STEM prompts related to the image. Then, a second layer evaluated the image, the prompt and solved the exercise without knowing the given answer. Finally, an automated layer moved forward the matches, or reserved the unmatches for a third revision.

The main project I have worked on in this area required an Image, a prompt, a ground truth answer, and an explanation. The model must have failed to solve STEM prompts related to the image. Then, a second layer evaluated the image, the prompt and solved the exercise without knowing the given answer. Finally, an automated layer moved forward the matches, or reserved the unmatches for a third revision.

2024 - 2025

Contributor

Scale AIComputer Code ProgrammingComputer Programming Coding

Fail and fix. Many projects in those months were about request to the model a difficult enough program that resulted in unsatisfactory code. The most extensive project I worked, had 3 model responses, where the best option should be chosen and improved. Evaluation and rating were also part of each task.

Fail and fix. Many projects in those months were about request to the model a difficult enough program that resulted in unsatisfactory code. The most extensive project I worked, had 3 model responses, where the best option should be chosen and improved. Evaluation and rating were also part of each task.

2024 - 2024

Education

Work History