Aaditya Kumar

Key Skills

Software

Appen

Labelbox

Scale AI

Top Subject Matter

No subject matter listed

Top Data Types

Computer Code Programming

Image

Text

Top Task Types

Computer Programming/Coding

Evaluation/Rating

RLHF

Freelancer Overview

I am currently pursuing engineering at NIT Srinagar and have been actively involved in training AI systems since November across multiple platforms, including Outlier, Soul, and Aligner. I have contributed to a variety of projects such as Reinforcement Learning with Human Feedback (RLHF), Supervised Fine-Tuning (SFT), API calling, and English data labeling. Through these experiences, I’ve gained valuable exposure to different aspects of AI training and model development, enhancing both my technical skills and understanding of real-world AI workflows.

Labeling Experience

INDEPENDENT CONTRACTOR

Scale AIComputer Code ProgrammingRLHFPrompt Response Writing SFT

Worked on cutting-edge AI training tasks focused on improving model behavior through Reinforcement Learning with Human Feedback (RLHF) and API integration. My contributions included: RLHF Evaluation: Assessed and ranked AI-generated responses based on alignment with human values, helpfulness, clarity, and factual correctness. Helped shape reward models used for aligning large language models with real-world user expectations. Multi-Turn Ranking: Evaluated complex, multi-turn conversations to ensure consistency, coherence, and high-quality responses across dialogue chains. API Calling Tasks: Designed and tested prompt structures for tool-augmented tasks where models interact with APIs. Ensured correct, context-aware API execution based on natural language input and task goals. Focused on crafting realistic and challenging scenarios that pushed models to demonstrate reasoning, tool usage, and response accuracy in dynamic contexts.

Worked on cutting-edge AI training tasks focused on improving model behavior through Reinforcement Learning with Human Feedback (RLHF) and API integration. My contributions included: RLHF Evaluation: Assessed and ranked AI-generated responses based on alignment with human values, helpfulness, clarity, and factual correctness. Helped shape reward models used for aligning large language models with real-world user expectations. Multi-Turn Ranking: Evaluated complex, multi-turn conversations to ensure consistency, coherence, and high-quality responses across dialogue chains. API Calling Tasks: Designed and tested prompt structures for tool-augmented tasks where models interact with APIs. Ensured correct, context-aware API execution based on natural language input and task goals. Focused on crafting realistic and challenging scenarios that pushed models to demonstrate reasoning, tool usage, and response accuracy in dynamic contexts.

2024

Education

Work History