Harsh Shukla

Key Skills

Software

Labelbox

OneForma

Remotasks

Telus

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

Prompt + Response Writing (SFT)

RLHF

Text Generation

Freelancer Overview

I am an expert in AI training. I have good hands on experience in helping train state-of-the-art (SOTA) Large Language Model (LLMs). I was acting as a domain expert in facilitating Reinforcement Learning via Human Feedback to Fine-tune AI models. I used to create Complex Prompts that can stump a model, improving AI responses, generating technical content, and reviewing the response to make it better. I played a crucial part in the Reinforcement Learning from Human Feedback (RLHF) process to fine-tune AI models.

Labeling Experience

Coding Prompt Writing and Response Rewriting

Scale AIComputer Code ProgrammingComputer Programming Coding

In this project, I taught the model to answer a variety of coding questions in a variety of languages so that it can perform well all over the world. I wrote coding prompt question. Al will respond to the prompt twice. Once the model generates a response to the prompt, review it carefully. Identify areas where the response fails short or makes mistakes. If one of the models does not have an issue, edit your prompt and try again until the model fails.

In this project, I taught the model to answer a variety of coding questions in a variety of languages so that it can perform well all over the world. I wrote coding prompt question. Al will respond to the prompt twice. Once the model generates a response to the prompt, review it carefully. Identify areas where the response fails short or makes mistakes. If one of the models does not have an issue, edit your prompt and try again until the model fails.

2024 - 2024

Image Generation and LLM Evaluation

Scale AIImageEvaluation RatingPrompt Response Writing SFT

This project was used to help train Al models to better understand and reason through Images across various disciplines. Challenge the model! Create image-based prompts across different difficulty levels, evaluate the model's responses, and add the correct final answer.

This project was used to help train Al models to better understand and reason through Images across various disciplines. Challenge the model! Create image-based prompts across different difficulty levels, evaluate the model's responses, and add the correct final answer.

2024 - 2024

Improve LLM by rating two or more responses on various rating dimensions.

Scale AITextRLHF

The work and inputs on this project will help improve a cutting edge large language model to provide more helpful, factually accurate and non-verbose responses. I achieved this by rating two responses from the model on various rating dimensions. The ratings are not based on "feels" but have a structured approach on arriving at a rating. It was a highly impactful project that will generate a dataset which improve current LLM capabilities.

The work and inputs on this project will help improve a cutting edge large language model to provide more helpful, factually accurate and non-verbose responses. I achieved this by rating two responses from the model on various rating dimensions. The ratings are not based on "feels" but have a structured approach on arriving at a rating. It was a highly impactful project that will generate a dataset which improve current LLM capabilities.

2024 - 2024

Education

Work History