Nykera Washington

Key Skills

Software

Data Annotation Tech

Scale AI

Other

Top Subject Matter

No subject matter listed

Top Data Types

Audio

Text

Image

Top Label Types

Emotion Recognition

Audio Recording

Transcription

Text Generation

RLHF

Fine Tuning

Evaluation Rating

Prompt Response Writing SFT

Question Answering

Text Summarization

Freelancer Overview

I am an experienced AI data trainer and annotation specialist with a strong background in developing high-quality training data for advanced AI models across multiple domains. My work has included creating and evaluating datasets for reasoning, writing, problem-solving, and structured analysis tasks, as well as performing voice acting and natural speech recording to support speech model development. I am skilled in assessing model outputs for accuracy, coherence, and safety, and providing detailed feedback to improve model performance. I adapt quickly to new tasks, and consistently meet quality benchmarks. My administrative and documentation experience further strengthens my attention to detail and ability to manage complex workflows in fast-paced, evolving environments.

Labeling Experience

Sparrowsignet

Scale AITextQuestion AnsweringText Generation

This project required participants to write a series of prompts and thorough rubrics according to instruction hierarchy and system steer-ability. The prompts had to align with a specific difficulty score (1 for basic everyday asks, 3 for specialized/niche information) and conflict with instruction requirements (no conflict, full conflict, or conflict with exception). The final prompt had to cause model failure in a specific area, such as instruction following, action, emotion, etc. The participants were also required to write very thorough rubrics that listed out every ask in the final prompt, both implicit and explicit, then score the final response based on that rubric, ensuring that the response scored under 70%.

This project required participants to write a series of prompts and thorough rubrics according to instruction hierarchy and system steer-ability. The prompts had to align with a specific difficulty score (1 for basic everyday asks, 3 for specialized/niche information) and conflict with instruction requirements (no conflict, full conflict, or conflict with exception). The final prompt had to cause model failure in a specific area, such as instruction following, action, emotion, etc. The participants were also required to write very thorough rubrics that listed out every ask in the final prompt, both implicit and explicit, then score the final response based on that rubric, ensuring that the response scored under 70%.

2025 - 2025

Citadel Choir

Scale AITextText GenerationEvaluation Rating

Required participants generate prompts, evaluate two model responses, and explain the pros and cons of each model response.

Required participants generate prompts, evaluate two model responses, and explain the pros and cons of each model response.

2025 - 2025

Melvins Mansion

Scale AITextQuestion AnsweringText Generation

This project required participants to write a series of prompts that led to model failure in a specific domain (action, tone, emotion, explicit instruction, etc) on the final response. Participants would also have to write thorough rubrics, outlining all implicit and explicit requirements of a perfect model response, score the model against that rubric, and then write a "golden response" that satisfied all the rubric criterion.

This project required participants to write a series of prompts that led to model failure in a specific domain (action, tone, emotion, explicit instruction, etc) on the final response. Participants would also have to write thorough rubrics, outlining all implicit and explicit requirements of a perfect model response, score the model against that rubric, and then write a "golden response" that satisfied all the rubric criterion.

2025 - 2025

Xylophone English Convo

Scale AIAudioEvaluation RatingAudio Recording

This project required two people to have a 10-15 minute conversation via Zoom to produce data that will train LLMs on natural speech patterns. I was a reviewer on this project, requiring me to listen to the full audio and score each recording based on the provided rubric. The rubric included fields like background noise level, audio artifact identification, speaker volume, conversation flow/naturalness, etc.

This project required two people to have a 10-15 minute conversation via Zoom to produce data that will train LLMs on natural speech patterns. I was a reviewer on this project, requiring me to listen to the full audio and score each recording based on the provided rubric. The rubric included fields like background noise level, audio artifact identification, speaker volume, conversation flow/naturalness, etc.

2025 - 2025

Multimodal with Rubrics

Scale AIImageQuestion AnsweringEvaluation Rating

This project required participants to write a prompt based on an image, then evaluate and compare two model responses for accuracy, interpretation of the image, and implicit/explicit requirements of the participant's written rubric based on their prompt.

This project required participants to write a prompt based on an image, then evaluate and compare two model responses for accuracy, interpretation of the image, and implicit/explicit requirements of the participant's written rubric based on their prompt.

2025 - 2025

Show more

Education

Work History