Rishabh Raj

Key Skills

Software

LabelImg

Label Studio

Scale AI

Other

Top Subject Matter

Generative AI

Computer Programming

LLM Prompt Engineering

Top Data Types

Computer Code Programming

Image

Text

Document

Top Task Types

Action Recognition

Bounding Box

Classification

Prompt + Response Writing (SFT)

Text Summarization

RLHF

Freelancer Overview

AI Data Trainer (Contract). Brings 3+ years of professional experience across legal operations, contract review, compliance, and structured analysis. Core strengths include Other. AI-training focus includes data types such as Computer Code, Programming, and Text and labeling workflows including RLHF, Evaluation, and Rating.

Labeling Experience

AI Data Trainer (Contract)

OtherRLHF

As an AI Data Trainer, I optimized the reasoning capabilities of large language models (LLMs) using Reinforcement Learning from Human Feedback (RLHF) and Supervised Fine-Tuning (SFT). I was responsible for evaluating complex code generations for technical accuracy and contributed to dataset creation for training purposes. This role required in-depth analysis of multi-turn logic and the minimization of model hallucinations. • Reviewed and rated model outputs in Python and Java. • Created labeled ground-truth datasets for code generation assessment. • Employed best practices in RLHF/SFT for AI safety and reliability. • Focused on multi-turn interactions and edge-case scenarios for robust LLM evaluation.

As an AI Data Trainer, I optimized the reasoning capabilities of large language models (LLMs) using Reinforcement Learning from Human Feedback (RLHF) and Supervised Fine-Tuning (SFT). I was responsible for evaluating complex code generations for technical accuracy and contributed to dataset creation for training purposes. This role required in-depth analysis of multi-turn logic and the minimization of model hallucinations. • Reviewed and rated model outputs in Python and Java. • Created labeled ground-truth datasets for code generation assessment. • Employed best practices in RLHF/SFT for AI safety and reliability. • Focused on multi-turn interactions and edge-case scenarios for robust LLM evaluation.

2024 - 2026

AI Content Specialist - Diamond Project

OtherText

As an AI Content Specialist on the Diamond Project, I led technical data labeling and localization workflows. I validated LLM responses against complex, domain-specific prompts to ensure project safety and utility alignment. My efforts contributed to achieving a 98% quality score for labeled data. • Performed data labeling and content validation for LLM-generated text. • Ensured adherence to strict safety and utility guidelines. • Oversaw technical and linguistic localization tasks. • Maintained quality and compliance across large-scale text datasets.

As an AI Content Specialist on the Diamond Project, I led technical data labeling and localization workflows. I validated LLM responses against complex, domain-specific prompts to ensure project safety and utility alignment. My efforts contributed to achieving a 98% quality score for labeled data. • Performed data labeling and content validation for LLM-generated text. • Ensured adherence to strict safety and utility guidelines. • Oversaw technical and linguistic localization tasks. • Maintained quality and compliance across large-scale text datasets.

2025 - 2025

God Garlic

Scale AITextText Generation

In this project, Our mission was to examine a prompt which will be divided into three main sections: System Prompt Developer Prompt User-Model Conversation The model will try to answer the user prompt using one or several tools provided within its structure using natural language as much as possible. There will be some cases when the model will use other tools to fulfill the user's requirement, and that's perfectly fine! The main goal is to focus on whether there's enough information in the prompt to meet the prompt's requirements.

In this project, Our mission was to examine a prompt which will be divided into three main sections: System Prompt Developer Prompt User-Model Conversation The model will try to answer the user prompt using one or several tools provided within its structure using natural language as much as possible. There will be some cases when the model will use other tools to fulfill the user's requirement, and that's perfectly fine! The main goal is to focus on whether there's enough information in the prompt to meet the prompt's requirements.

2024 - 2024

Education

Work History