Evancarson Jesejeri

Key Skills

Software

AWS SageMaker

Appen

Google Cloud Vertex AI

Lionbridge

Remotasks

Sama

Scale AI

Top Subject Matter

No subject matter listed

Top Data Types

Audio

Computer Code Programming

Image

Top Task Types

Audio Recording

Computer Programming Coding

Fine Tuning

Prompt Response Writing SFT

RLHF

Freelancer Overview

I have over 5 years of experience in AI training and data labeling, working extensively with large-scale datasets to train and fine-tune state-of-the-art models, including large language models (LLMs). My work has involved reinforcement learning from human feedback (RLHF), prompt evaluation, conversational AI tuning, and quality control for AI-generated content. I’ve collaborated on projects that required high-precision annotation, ranking model responses for coherence, factual accuracy, bias, and tone. My hands-on experience with tools like Label Studio, Appen, Scale AI, and proprietary platforms at Google DeepMind has allowed me to develop deep insight into the human-in-the-loop AI training pipeline. What sets me apart is my ability to blend technical fluency with linguistic and contextual sensitivity—ensuring not just accuracy in annotations, but also ethical and user-aligned outputs. I have a strong background in NLP, data quality assessment, and multilingual prompt review. Whether working on LLM alignment, toxicity filtering, or dialog flow optimization, I bring a meticulous, feedback-driven approach that enhances model performance and trustworthiness.

Labeling Experience

LLM Response Ranking and Prompt Evaluation for Conversational AI

AppenImageClassificationText Generation

This project focused on the human evaluation and tuning of large language model (LLM) outputs using Reinforcement Learning from Human Feedback (RLHF). My role involved ranking AI-generated responses based on accuracy, coherence, tone, and relevance to user prompts. I also performed prompt+response crafting for supervised fine-tuning (SFT) and flagged outputs that were biased, toxic, or misleading. The data was primarily in English, with some exposure to multilingual prompts. Over the course of the project, I annotated and evaluated over 10,000 data points, following strict quality control protocols and review guidelines set by the model alignment team. Tools used included Appen, Scale AI, and proprietary platforms developed by DeepMind.

This project focused on the human evaluation and tuning of large language model (LLM) outputs using Reinforcement Learning from Human Feedback (RLHF). My role involved ranking AI-generated responses based on accuracy, coherence, tone, and relevance to user prompts. I also performed prompt+response crafting for supervised fine-tuning (SFT) and flagged outputs that were biased, toxic, or misleading. The data was primarily in English, with some exposure to multilingual prompts. Over the course of the project, I annotated and evaluated over 10,000 data points, following strict quality control protocols and review guidelines set by the model alignment team. Tools used included Appen, Scale AI, and proprietary platforms developed by DeepMind.

2020 - 2025

Education

Work History