Aditya Raj

Key Skills

Software

Labelbox

Other

Scale AI

Top Subject Matter

No subject matter listed

Top Data Types

Computer Code Programming

Video

Top Label Types

RLHF

Function Calling

Prompt Response Writing SFT

Evaluation Rating

Computer Programming Coding

Object Detection

Action Recognition

Tracking

Freelancer Overview

I have hands-on experience in AI model training and evaluation, with a strong foundation in data annotation and quality benchmarking. At Turing AI, I contributed to the development and optimization of AI reasoning models, participating in supervised fine-tuning (SFT) and reinforcement learning from human feedback (RLHF) pipelines to enhance model accuracy and alignment. My academic projects include building a yoga pose correction web app using machine learning and computer vision, where I utilized TensorFlow.js and MediaPipe for real-time pose detection and feedback. I am skilled in C++, Python, SQL, and JavaScript, and have a solid understanding of data structures, algorithms, and database management. My experience with both the technical and evaluative aspects of AI systems enables me to ensure high-quality training data and robust model performance.

Labeling Experience

LLM Python and C++ Trainer

LabelboxComputer Code ProgrammingRLHFFunction Calling

The primary scope of the project was to train an Anthropic model for code evaluation by delivering clear, structured, and high-quality feedback. This involved continuously providing well-designed prompts and detailed evaluations to help the model compare, refine, and improve code outputs. The objective was to ensure that, over time, the model could consistently generate production-ready code that reflects the standards, reasoning, and best practices of a senior-level software developer.

The primary scope of the project was to train an Anthropic model for code evaluation by delivering clear, structured, and high-quality feedback. This involved continuously providing well-designed prompts and detailed evaluations to help the model compare, refine, and improve code outputs. The objective was to ensure that, over time, the model could consistently generate production-ready code that reflects the standards, reasoning, and best practices of a senior-level software developer.

2025 - 2025

Video Annotator

OtherVideoObject DetectionAction Recognition

I worked as a Video Annotator on an AI model training project focused on improving video understanding and evaluation. My responsibilities included annotating and reviewing video data using spatio-temporal labeling techniques such as action recognition, object tracking, and frame-level evaluation. I ensured annotations were accurate, consistent, and aligned with project guidelines to support high-quality training data. In addition to annotation, I evaluated model outputs and provided structured feedback to help improve prediction accuracy and overall model performance. This work contributed to training and refining machine learning models for reliable video analysis and decision-making.

I worked as a Video Annotator on an AI model training project focused on improving video understanding and evaluation. My responsibilities included annotating and reviewing video data using spatio-temporal labeling techniques such as action recognition, object tracking, and frame-level evaluation. I ensured annotations were accurate, consistent, and aligned with project guidelines to support high-quality training data. In addition to annotation, I evaluated model outputs and provided structured feedback to help improve prediction accuracy and overall model performance. This work contributed to training and refining machine learning models for reliable video analysis and decision-making.

2025 - 2025

Prompt writer and evaluater

Scale AIComputer Code ProgrammingPrompt Response Writing SFT

Here is an enhanced, professional version with improved clarity and impact: I was assigned to design high-quality prompts across STEM and Computer Science domains and to provide detailed, structured feedback on model responses. This included evaluating the accuracy, reasoning depth, and technical correctness of outputs, and refining prompts to improve the model’s understanding, problem-solving ability, and response quality over successive iterations.

Here is an enhanced, professional version with improved clarity and impact: I was assigned to design high-quality prompts across STEM and Computer Science domains and to provide detailed, structured feedback on model responses. This included evaluating the accuracy, reasoning depth, and technical correctness of outputs, and refining prompts to improve the model’s understanding, problem-solving ability, and response quality over successive iterations.

2025 - 2025

Codeforces Problem tutor

OtherComputer Code ProgrammingRLHFEvaluation Rating

I was responsible for designing Codeforces-style programming questions that the Qwen model was initially unable to solve. My task involved analyzing the model’s limitations, then providing a clear, step-by-step solution that explained the underlying logic and approach in detail. Finally, I delivered a complete, optimized reference implementation to train the model, enabling it to understand the problem-solving process and improve its ability to handle similar competitive programming challenges independently.

I was responsible for designing Codeforces-style programming questions that the Qwen model was initially unable to solve. My task involved analyzing the model’s limitations, then providing a clear, step-by-step solution that explained the underlying logic and approach in detail. Finally, I delivered a complete, optimized reference implementation to train the model, enabling it to understand the problem-solving process and improve its ability to handle similar competitive programming challenges independently.

2025 - 2025

Education

Work History